Final Report Summary - PHARMAS (Ecological and human health risk assessments of antibiotics and anti-cancer drugs found in the environment.)
Executive Summary:
* The project focussed on two categories of pharmaceuticals: antibiotics and anticancer drugs.
* All available information about a selection of these pharmaceuticals was entered into a database.
* Modelling was used to predict the likely concentrations of both antibiotic and anticancer drugs in rivers throughout the European Union; predicted concentrations were in the ng/litre range.
* Measurement of actual concentrations in both surface and drinking water confirmed that concentrations were low, and sometimes not detectable.
* Food, as well as water, was identified as a potential route of exposure to people.
* Toxic effects in a range of aquatic organisms occurred only at high concentrations of cytotoxic drugs - well above environmental concentrations.
* Theoretical work suggested that because many different antibiotics and anticancer drugs are present simultaneously in the environment, then the risk they pose should be based on realistic mixture scenarios.
* Wastewater treatment technologies were shown to degrade parent pharmaceuticals into a range of transformation products, many of which could not be identified or their potential toxicities assessed.
* Probabilistic risk assessment techniques were employed to identify locations in Europe showing the highest human and ecological risks due to the presence of antibiotics and anticancer drugs.
* A detailed qualitative socio-empirical study of the requirements of different stakeholders identified the information they sought.
* Although lack of appropriate data was identified as a major constraint, nevertheless a prototype for an EU-wide risk and hazard classification system for pharmaceuticals was developed.
* It was concluded that the cytotoxic drugs studied are unlikely to pose a serious threat to the environment. The situation is less clear for antibiotics, due to the concern about induced antibiotic resistance in microbes and the potential transfer of this resistance to pathogens.
* Human health is unlikely to be affected adversely by environmental exposure to the drugs studied.
Project Context and Objectives:
Pharmaceutical products (PPs) in the environment represent an increasing concern for environmental scientists as well as regulatory institutions. Some PPs and their metabolites are not removed during conventional biological treatment and enter the water supply via wastewater-treatment plants (WWTPs) and can also reach drinking waters. Much has been published concerning the occurrence, fate and behaviour of PPs, but risk assessment for the environment and for human health still remains weakly investigated.
According to the international literature, it appears that most of the up to now investigated pharmaceuticals do not pose an acute threat to the environment, but that a small number do, and that none appear likely to pose a significant threat to human health (via environmental exposure). These conclusions are supported by a recent and detailed human health risk assessment (Cunningham et al., 2009 ). However, this apparent certainty that there is no human health risk posed by the presence of human pharmaceuticals in the environment is not supported by deeper thinking. For example, the apparently reassuring risk assessments of Cunningham et al (2009) are based on potential risks to healthy adults: the more sensitive and hence vulnerable foetus, children, the elderly and the infirm were not considered. The real situation is that, currently, there are many more uncertainties than certainties, which leaves scientists, the public and the press still unconvinced that drinking water containing a tiny quantity of a pharmaceutical is completely harmless.
The Pharmas project was launched in this context by adopting an interdisciplinary approach to improve the knowledge on risks for human health and ecosystems, due to exposure to pharmaceuticals and their derivates which are released into the environment. An additional key point of the proposal is to include the main transformation products (metabolites, degradation and treatment by products) with the target molecules. The characterisation of such risk has to be unquestionable, robust and reliable. Furthermore, for decision makers and also for the general public, it is important to make this science easily understandable and available.
As it would obviously be impractical to study all pharmaceuticals, or even a reasonable proportion of them, Pharmas chose to investigate antibiotics and anticancer drugs because they are considered most likely to be a threat to the environment and/or human health. Anticancer drugs can be cytotoxic, genotoxic, mutagenic and teratogenic, and it is generally accepted that no threshold of safety can be given for some of them. Moreover, this group of pharmaceuticals has been studied very little. Concerning antibiotics, although their presence in the environment has already been well studied, the extent to which they are responsible for the induction of genetic resistance is still not well understood.
The final objective is to reduce the level of uncertainty in the current risk assessment procedure. PHARMAS develops a general procedure taking into account scenarios of exposure, as well as the impact of the chosen PPs on organisms, in order to define the range of possible risks for these populations. The project also covers toxicity of mixtures and transformation products. Moreover, social and policy considerations have been added in order to, respectively, inform the public and to support appropriate changes in regulation.
The Pharmas consortium consists of 13 partners from a total of 6 European countries. The project employs a multidisciplinary approach to the problems associated with PPs in the aquatic environment. The partnership involves partners with complementary skills and competencies. This consortium has been designed to meet the research objectives of this collaborative project.
The project was planned over a period of 36 months, the expected results complying with the global objectives of the project were:
- to determine the human and animal exposure to the target molecules by measuring and modelling the concentrations found from the resource (mainly surface water) to the tap (drinking water)
- to provide the missing data on ecotoxicological and toxicological thresholds of concern, by the implementation of targeted ecotoxicological experiments, the compilation of all available literature and an assessment of the sensitivity of particular populations (embryos/newborns)
- to investigate the toxicity of realistic mixtures involving antibiotics and anti-cancer drugs in several dedicated case studies
- to produce probabilistic estimates of the risk caused by the exposure of wildlife and humans by analyzing the uncertainty and variability in the exposure estimates
- to evaluate the usefulness of ecological effect data for the assessment of human effects and to explore options to develop a common effect endpoint for human and ecological risks
- to identify stable transformation products (through treatment processes) and derive very first knowledge how to render future pharmaceuticals better biodegradable in the aquatic environment
- to develop a web-based system primarily designed as an easily accessible decision support tool for practitioners (e.g. physicians and pharmacists) providing straightforward action alternatives
- to study perception of the topic and improve information to the public
- to stimulate discussion by all stakeholders and beneficiaries of the work by supporting communication and collaboration between scientists and regulators
- to contribute to an evolution of the regulation and EU policies by distilling the policy-relevant information, products and conclusions of the project for the relevant policy fields
To help Pharmas in its objectives and works, the project was structured (Figure 1) around 8 workpackages: 6 technical, 1 for dissemination and 1 for coordination. An Advisory Group, composed of a panel of experts, representative of several stakeholders, has been formed to comment and improve the strategy of the project and its results.
Figure 1: Pharmas organisation
Project Co-ordination
The project overall ran extremely smoothly, with excellent co-operation between partners and frequent opportunities to meet in person and discuss any issues of concern. Project meetings were held annually where at least one representative of each partner was present on all occasions. In addition (sometimes in conjunction with the annual meetings), satellite meetings were held sometimes to discuss particular WPs and sometimes with all WP leaders (the ‘scientific council’) present. A separate ‘Advisory Group’ meeting was also convened to discuss stakeholder issues, at which all of the regular Pharmas stakeholders were present as well as several additional interested parties (including Helen Clayton, DG Env) [see deliverable 8.3]. Thus we attempted to maintain full communication links with stakeholders and to encourage them to present their views on the project.
The scientific council was composed of the workpackage leaders and aimed to manage the scientific activities of the project. Six scientific council meetings were organized during the project: February 21-22rd 2011 (at the KoM), November 25th 2011, February 9-10th 2012 (at the annual meeting), May 23rd 2012, February18-19th 2013 (at the second annual meeting) and March 11-12th 2014 (at the final meeting). The kick-off meeting was a good opportunity to launch the project, to establish the mode of functioning of the project hoped for by the coordinator and to define the role of WP leaders; the annual meetings and final meetings allowing to overview the results of the project and to discuss interactions with other WPs. At these meetings, WP leaders also presented their WPs to the advisory group. The other two Scientific Council meetings involved WP leaders only, and allowed discussion on work progress, administrative and financial aspects, and to envisage future works. In no case was a delay in a specific deliverable thought to cause a negative impact on the progress of other areas of the project.
Project Results:
C. A description of the main S&T results/foregrounds
NB/ Figures referred to in this section can be found in the attached pdf file which contains the text and figures of the entire report prepared by the consortium.
Pharmaceuticals in the Environment raise a number of questions about the risks they represent for wildlife and humans. Currently, the risk assessment approaches are not able to give us a strong and definitive answer because many uncertainties still exist. In the Pharmas project, the workpackages were designed to improve this situation by gathering information on exposure, effects (single and mixture), presence and effect of transformation products for the compounds of greatest concern. The following section reports the main results obtained during the project.
The project was focussed on two categories of molecules of high concern: antibiotics and anticancer drugs. Among the wide amount of such substances, a selection of 19 molecules was made, based on the usage, the available literature and the analytical possibilities:
- Anticancer drugs: 5-Fluorouracil (5FU), Capecitabine (CAP), Ifosfamide (IFO), Cyclophosphamide (CPA), Hydroxycarbamide (HYD), Carboplatin (CAR), Methotrexate (MTX), Tamoxifen (TAM) + Imatinib was at least in WP5 investigated as an example of a new drug, see also C1.1.2
- Antibiotics: Chlortetracycline, Doxycycline, Oxytetracycline, Tetracycline (TET), Cefuroxim, Ceftriaxone, Trimethoprim (TMP), Erythromycin (ERY), Ciprofloxacin (CIP), Levofloxacin, Ofloxacin.
All the available data concerning these molecules were collected from the scientific literature, medical/phamaceutical literature, at both national and international levels but also from available data bases (MISTRA, KNAPPE data base, ERAPHAR database, pubchem substance, ChemBank, drug product data base of Health Canada …) and from the manufacturers.. They included : physico-chemical properties (solubility; adsorption; structure), behaviour in humans and animals (metabolism data, dose-response data in various species, preferably a selection of representative aquatic species (algae, Daphnia, fish, etc) and rodents (for extrapolation to humans)), data on mode(s)-of-action, data on potential interactions, contraindications, characteristics of usage by different countries (prescription; consumption; therapeutic uses), environmental fate and behaviour (emission, occurrence; degradation, transfer) and they are described in detail in the deliverable 1.1.
C.1. Exposure assessment:
Calculation of exposure is a determinant element for a reliable risk assessment for both human and environmental ecosystems. Modelling and real measurement have been implemented in the project on several of the molecules of interest.
C.1.1. Modelling
In order to consider several way of exposure, modelling was performed on water but also on foodstuff.
C.1.1.1. Modelling environmental concentrations of exposure via drinking water
In a first step, the work focused on the potential concentrations expected in sewage effluent and surface waters and established the final predicted values for the cytostatic drugs (5FU, CAP, CP, and CAR) and the antibiotics (CIP, SUF, TRI and ERY) for the different countries in the European Union. To examine potential concentrations of these chemicals throughout European surface waters, the geographic-based water resources model GWAVA run in a water quality mode was used. This model uses geographic data on the location and size of the human European population and their association with STPs. The version of GWAVA used here incorporates a newly available and extensive dataset (2009-2010 information) of locations and number of people connected to sewage discharge points in Europe. Information used in the model included flows through STPs, the hydrology, drug consumption, excretion, removal in sewage, and in-stream half- life, ... The model calculates the water concentrations of chemicals throughout European water courses using an overlay of 177,470 grid squares (cells) of approximately 6 x 9 km (5 by 5 Arc minutes) dimensions. Details of each parameter are further explained in deliverable 1.2.
There are uncertainties in the model parameters determining effluent concentrations, which are critical in estimating river concentrations. In order to assess the impact of this uncertainty, a series of scenarios were run to establish the range of likely river concentrations (and hence likely EQS exceedence), based on the reported literature values. These scenarios were a best case - low excretion, high sewage removal and high in stream dissipation; a worst case - high excretion, low removal and slow in stream dissipation and an expected case, which used weighted average values for these parameters. The expected scenario based on the mean literature values gave reasonable agreement with the few available measured data (detail in the deliverable 1.3). The good correlation between values predicted in this study and those observed in sewage effluent are encouraging and further refinement of the model will be necessary in the future as more information on the fate and behaviour of these compounds becomes available.
Figure 2 presents an example of predicted concentrations for cyclophosphamide in European rivers.
Figure 2: Predicted cyclophosphamide (CP) concentrations in surface water based on mean excretion rate, mean sewage treatment removal, and 50%ile flow across the European Continent taking into account differing national per capita consumption and wastewater discharge values from the GWAVA model.
Considering the other cytostatics, there was a surprising difference in the popularity of these drugs across European nations. The predicted mean effluent concentrations ranged from 2 to 40 ng/L for CP, 0.8 to 2.5 ng/L for carboplatin, 0.3-2.5 ng/L for 5FU, and 8.5-87 ng/L for capecitabine. By linking with the geographic based water quality model it is expected that the majority of European rivers would have concentrations below 1 ng/L for these cytostatic drugs that are considerably below concentrations so far reported to have effects on aquatic wildlife. There does not appear to be any widespread threat to European aquatic wildlife based on our current knowledge, because even in the 90%ile prediction around 80% of European surface waters predicted concentrations for these drugs of (generally) below 0.1 ng/L. The issue of water abstraction for drinking water and foetal health may still require further research. Given its potentially high effluent concentrations and good oral absorption by humans, capecitabine certainly seems worthy of further environmental research.
The GWAVA model was also used for antibiotic modelling in surface water. The two scenarios modelled were expected and best case. The first task was to provide a concentration map for each compound using only the best case scenario (annual average). Figure 3 illustrates the case of Trimethoprim.
Figure 3: Predicted trimethoprin (TRI) concentrations in surface water based on best case scenario (lowest excretion rate and highest sewage treatment removal) and mean flow across the European Continent taking into account differing national per capita consumption and wastewater discharge values from the GWAVA model (little used in Spain).
With the expected scenario at the highest end of results (90%ile) antibiotic concentrations could reach 10s to 100s of ng/L in 10% of river lengths (see deliverable). If we assumed the worst case situation of high excretion, poor removal and low flow then concentrations could reach 100 ng/L in up to 10% of river lengths for all the antibiotics except ciprofloxacin. The most toxic of the antibiotics appears to be ciprofloxacin but under most circumstances all European rivers are around 2 orders of magnitude below the effect level. Trimethoprin appears to be of the least concern having around 6 orders of magnitude safety factor. We cannot comment on to what extent such river concentrations might stimulate antibiotic resistance.
The results of GWAVA models for several European countries is reported in the deliverable 1.4
C.1.1.2. Identification of possible routes of exposure via foodstuffs
Drinking water is not the only route of exposure of human to pharmaceutical products. Food was also investigated as potential vector of contamination.
Indirect human exposure was assessed via the exposure routes from wastewater to rivers and to agricultural soils, and further via bioaccumulation in fish and crops to humans (Figure 4).
Figure 4: Emission pathways and environmental compartments considered in the study.
The model Activity Simple Treat was used. This is a version of Simple Treat adapted to deal with ionisable chemicals. The stated variable is not concentration nor fugacity, but the activity of the chemical. A dynamic plant uptake model for ionisable compounds was developed, which combines the standard model approaches with the concept of the cell model. To validate this model, a literature search for measured concentrations in plants has been undertaken.
The chemical data set (antibiotics and cytostatic drugs) included polar neutral substances, mono- and bivalent acids, mono- and bivalent bases, zwitterions of various types, and metallorganics. This made chemodynamic calculations challenging. A major effort was to modify approaches, among them SimpleTreat, regressions for bioconcentration in fish, and dynamic models for transport of water and substances in soil and plant, plant uptake models, so that multispecies calculations for the various ionic species could be undertaken. For the bioaccumulation in fish and plants, varieties of the cell model for each type of electrolytes were established because existing regressions for the bioconcentration (BCF) are not applicable for many of the investigated compounds. Usual models focus on partitioning between lipids and water. That process is of little relevance for the very polar medical drugs, such as zwitterionic antibiotics (for example, ciprofloxacin, tetracyclines). As novelty we added the adsorption to proteins into the models. The model system was consistently parameterized for the Danish scenario (but can be adapted to any other European scenario).
The results (Table 1) indicate that amendment of agricultural fields with sewage sludge and the subsequent translocation into food crops is most relevant for indirect human environmental exposure to pharmaceuticals, more than ingestion of (river) fish. Highest dietary exposure with root crops (incl. potato), leafy vegetables and fruits (incl. cereals) was predicted for the antibiotics tetracycline, ciprofloxacin, trimethoprim, and others and for the anticancer drug capecitabine and imatinib (all > 1 g/d). Plants generally concentrate polar and nonvolatile compounds (if they are sufficiently persistent to reach the plant), while fish accumulate more lipophilic substances. Most of the investigated compounds were ionizable and thus very polar, only tamoxifen belongs to the very lipophilic compounds.
Table 1. Dietary uptake of antibiotics by women in Denmark, 50%-percentile. Unit ng/d.
women ng/d root crops leafy vegetables fruits & cereals sum crops fish
Ceftriaxone 0.24 0.04 0.01 0.30 0.16
Cefuroxime 0.00 0.00 0.00 0.00 2.3
Chlortetracycline 0.00 0.00 0.00 0.00 0.00
Ciprofloxacin 175.6 38.8 21.4 235.7 7.5
Doxycycline 301.9 14.8 7.3 324.0 53.6
Erythromycin 37.9 156.4 29.5 223.8 0.6
Levofloxacin 34.9 16.9 7.2 59.0 5.9
Ofloxacin 7.5 0.3 0.4 8.3 2.2
Oxytetracycline 0.65 0.16 0.11 0.91 0.22
Tetracycline 650.4 5.0 1.2 656.5 36.2
Trimethoprim 158.3 841.7 164.3 1164.3 5.15
A comparison of predicted to measured concentrations in soil and plant was satisfactory in most cases. However, the variation of measured data is quite high. The BCFs (bioconcentration factors) leaves to soil of trimethoprim, erythromycin and ceftriaxon are above 1 kg:kg. Accumulation in corn is one to two orders of amount lower. Practically no comparison to experimental data could be done for the bioconcentration in fish. Therefore, the results possess therefore high uncertainty and must be considered as preliminary.
In summary, it appears that dietary exposure to human antibiotics via indirect environmental exposure is higher for the pathway sludge - soil - crops than for the pathway effluent - river – fish. In addition by using Danish data, the dietary exposure to human antibiotics via indirect environmental exposure was shown to be particularly high (close to 1 g/d) for several antibiotics, namely the compounds trimethoprim, tetracycline, ciprofloxacin, doxycycline, erythromycin, and a few anticancer drugs (imatinib, capecitabine). However, there remain considerable uncertainties. The deliverable 1.5 details the whole works of this section.
C.1.2. Measured Environmental Concentration
Part of the project was dedicated to real measurement of the chemicals of concern in various water matrices, in order to evaluate the degree of contamination of antibiotics and anticancer drugs, in comparison to other pharmaceutical products.
In a first step an interlaboratory exercise on ATB was organised. 14 laboratories participated including 2 members of the Pharmas consortium. Considering the different profiles of the participants (obtained by a questionnaire) and the capacity of the host laboratory to produce the reference materials, the following antibiotics were chosen: Ofloxacin, Ciprofloxacin, Sulfamethoxazole, Trimethoprim, Erythromycin. 2 matrices were investigated: surface and drinking water. Tests of stability, preservation and transport conditions have been performed. No standard protocol was imposed, neither for the extraction/concentration step nor for the analytical methods/devices.
A total of 77 samples were analyzed to determine concentrations of selected antibiotics and 611 results (including parallels, excluding
In a second step, analytical campaigns were performed. More precisely, several studies were performed in France and The Netherlands that aimed to evaluate the level of contamination of both raw (resources used to produce drinking water) and drinking waters and consequently the level of exposure of the population. Several analytical measurements were performed usingdrinking waters and their resources, by the laboratories of the consortium. Two series of studies were done in France and in The Netherlands.
A selection of target molecules was analysed, which include antibiotics and anticancer drugs selected in the project but also additional antibiotics and other pharmaceuticals (included in the multicomponent analytical methodologies employed). Table 2 presents the molecules analysed for this task.
Table 2: List of pharmaceuticals analysed
Molecules of interest for the project:
Ciprofloxacin Chlortetracyclin Erythromycin Ofloxacin
Oxytetracyclin Sulfamethoxazole Tetracyclin Trimethoprim
Cyclophosphamide Imatinib 5FU
Additional antibiotics:
Danofloxacin Difloxacin Enoxacin Enrofloxacin
Flumequin Lomefloxacin Nalidixic Acid Norfloxacin
Oxolinic Acid Pipemidic Acid Sulfamethazin Ofloxacin+Levofloxacin
Sarafloxacin
Additional molecules:
Atenolol Cafeine Carbamazepine Codeine
Iopromide Morphine Oxazepam
Sorafenib Erlotinib Sunitinib
Results from France showed only 1 positive result – antibiotic at 17 ng/L - out of more than 500 analyses. Therefore, considering the exposure dose calculation and the concentration found in drinking water, it can be concluded that human exposure to antibiotics and anticancer drugs via drinking water is unlikely. Considering other pharmaceutical products (PPs) analyzed in this work, the maximum concentration in drinking water was found to be 37 ng/L for Iopromide. Even present in drinking water (DW), such concentrations corresponds to an exposure to 1 ng/ kg bw/ d, 3.7 ng/ kg bw/ d, 5.5 ng/ kg bw/ d respectively for adults, children and infants.
Concerning experiments undertaken in the Netherlands, a smaller survey was carried out including surface water, groundwater and drinking water samples on a limited number of targets: 3 classical antibiotics and 5 anticancer drugs including 4 protein kinase inhibitors (PKIs) (Imatinib, Sorafenib, Sunitib and Erlotinib). The concentrations measured were found to be higher in sewage effluent than in ground and drinking water. Maximum concentration for antibiotics in surface water was quantified at 19 ng/L. PKIs were measured in surface water and even in drinking water but in the last case at concentration below 1 ng/L. Deliverable 1.7 reports in detail the results of these analytical tasks.
C.2. Effect assessment
For many of pharmaceuticals detected in the aquatic environment, the possible impacts on the environment and public health are unclear. Indeed, very few data on ecotoxicology are currently in the public arena, either for single substances (in particular for cytotoxic drugs), or mixtures.
This part of the project aimed to provide some answers as to whether these drugs might be of concern at the concentrations at which they are detected in the environment.
It seems to be widely accepted that environmental exposure of humans to pharmaceuticals (in drinking water, for example) will have no effects. This is because the exposure dose (in food or drinking water) is likely to be far lower than the therapeutic dose (the one given to patients). But in reality there is very little information to enable us to either confirm or refute this hypothesis. One concern is that there may be particularly sensitive sub-groups of humans (e.g. foetuses, or the elderly and infirm people).
In (probable) contrast to humans, it is known that environmental exposure to pharmaceuticals can adversely affect wildlife. Nonetheless, at the time of writing the Pharmas proposal, essentially little is known about possible effects of antibiotics and anti-cancer drugs on aquatic organisms. We aimed to fill that knowledge gap with a focus on the effects of anti-cancer drugs on fish, invertebrates, and algae.
Finally, concern has been expressed that the human foetus and/or young child may be more sensitive than the adults of the species. Although we clearly cannot test this hypothesis using human subjects, the project aimed to compare the sensitivities of young and adult fish. Fish are often good surrogates for mammals (see, for example, Zelikoff, 1998 ; Bols et al, 2005 ), so we anticipate that the results can be extrapolated to human health (for risk assessment purposes).
In this context, works on single substances were to (i) determine the minimum doses of anticancer drugs that cause measurable effects on humans with particular emphasis on sensitive sub-groups such as the foetus, (ii) assess the genotoxic effects of some anticancer drugs on fish, an aquatic invertebrate and algae, and (iii) compare the sensitivities of embryo and adult fish to some anticancer drugs.
The assessment of the effects of cytotoxic drugs in humans was not easy to perform due to the limited data available at public area. However, the topic was addressed and reported in deliverable 2.1. In that report, the main conclusions were:
- For different pharmaceuticals the difference between the daily therapeutic dose and the likely level of intake via drinking water and/or fish consumption ranges from two to ten or more orders of magnitude.
- All assessments published to date have concluded that the risk to human health from the presence of pharmaceuticals in the environment is very low, if not negligible.
- There are some uncertainties associated with this conclusion of very little or no risk, but these are unlikely to alter that conclusion.
- Particularly sensitive sub-groups of people may exist, but even these are unlikely to be adversely affected by the presence of human pharmaceuticals in the environment.
- Cytotoxic drugs are a special case because they are designed to be toxic. Although environmental concentrations of these drugs are currently largely unknown, they are very unlikely to be high enough to cause adverse effects to people inadvertently exposed to them.
- Reassurance for the conclusion of little or no risk to human health is provided by the fact that the health of the foetus is not affected if its mother undergoes chemotherapy for cancer during the second or third trimester of pregnancy (see a recent report on this subject by Amant et al, 2012).
The potential for effects of environmental exposure to cytotoxins on foetal health during the first trimester of pregnancy are less clear, but given that :
a) there are expected to be only very low concentrations (at the absolute most in the ng/L range) of cytotoxic drugs present even in surface water;
b) that concentrations in treated drinking water are likely to be lower still (and were undetectable in most of the samples analysed for this project – see above conclusions on measured and predicted concentrations of these compounds in surface and drinking water);
c) that even the most potent cytotoxin tested here (cisplatin) only induced toxic effects at several hundred µg/L;
- we think that the exposure levels via drinking water even to this vulnerable group will not be at concentrations high enough to induce adverse effects.
With particular reference to the potential for increased sensitivity of human sub-groups (e.g. exposure of the foetus during pregnancy, or the elderly/infirm), it was concluded that although an extremely ill person might react badly to a pharmaceutical product prescribed to them, it seems unlikely that amounts present in drinking water would be sufficient to trigger adverse effects. Further, a recent study (Amant et. al., 2012 ) has shown no effects on foetuses of pregnant mothers administered chemotherapy during the second or third trimester of pregnancy, suggesting that the much lower doses of cytotoxic drugs present in drinking water are very unlikely to affect the unborn foetus.
The assessment of genotoxic effects was investigated. Although a number of cytotoxins were assessed in some assays, the main focus was on three drugs, namely 5-Flourouracil (5-FU), cyclophosphamide (CPA) and cisplatin (CIS). Two immunocytochemistry assays (using antibodies to H2AX and RAD51, which are DNA repair enzymes) were originally selected for development in fish and algae, but they were found to be unreliable and unrepeatable using the tissues collected. A considerable amount of time was spent trying to develop this method, because it was felt that if a reliable assay had been forthcoming it would have provided invaluable and unique information on the toxic effects of these drugs in fish. As it was unsuccessful, it was decided to focus on the toxic effects of cytotoxins in algae, daphnia and adult zebrafish. A considerable amount of data describing the toxicity of these drugs in zebrafish embryos was produced. In addition, the measurement of the expression levels of genes encoding DNA repair enzymes (specifically, RAD51), and also the tumour suppressor gene P53, (which can be activated in response to DNA damage) in exposed zebrafish was performed in both adults and embryos.
In general, ecotoxic effects of cytotoxic drugs in all organisms studied (algae, daphnia, zebrafish adults and zebrafish embryos) were found to occur only at extremely high concentrations (several hundred micrograms or even several milligrams per litre), far higher than would be observed in European aquatic environments. This does not eliminate the potential for ecotoxic effects in areas where mixtures of these chemicals are present, but it does suggest that individually these drugs are less likely to cause harm to aquatic organisms, than we might have expected given their known toxic nature, particularly at the concentrations predicted to be found in surface waters, (albeit that toxic doses used in humans are at higher concentrations than those used in this project).
Cisplatin was found to be (relatively speaking) the most potent toxin in both the daphnia and zebrafish embryo assays. This compound was also tested in a pilot study at a single concentration alongside 5-FU and cyclophosphamide using adult zebrafish, however the cost of cisplatin excluded it from a comprehensive dose-response study with adult zebrafish (Partner 1 uses a flow-through exposure system, so several thousand euros worth of the chemical would have been required to conduct a study using this system). Nevertheless, this compound was only toxic in the assays performed here at concentrations exceeding those reported in the environment; a concentration of several hundred µg/L cisplatin was required to elicit a response in the zebrafish embryo toxicity assay, and in the daphnia immobilisation assay the EC50 of this chemical was 2.06 mg/L.
Algal inhibition assay
Inhibition of growth of algae (Pseudokirchneriella subcapitata) was assessed using kits purchased from MicroBioTests, which enable the calculation of a 72 hr EC50. Briefly, test organisms (Pseudokircheniella subcapitata) were stored immobilised in ‘algal beads’ and were de-immobilised on the first day of the assay. Algae were incubated in medium containing the toxicant at 22±1°C, under constant uniform illumination, for 72 hours. OD was determined using spectrophotometer after 24, 48 and 72 hours incubation. The three cytotoxins selected for use in this assay were found to have relatively high EC50 values in this assay (i.e. they are not considered to be highly toxic, and do not show effects in this assay at levels observed in the aquatic environment); these values concur with those reported by Zounkova (2007) in a similar algal growth inhibition assay (both in terms of rank order of potency and, broadly speaking, absolute potency). The EC50 values of these compounds were as follows: 5FU: 0.44 mg/L; CIS: 7.4 mg/L; CPA:3179 mg/L (see Figure 5).
Figure 5: Results of the algal inhibition assay undertaken with cytotoxic drugs
Daphnia immobilisation assay
Daphnia immobilisation was assessed using kits also purchased from MicroBioTests. Briefly, ephippia of Daphnia magna were hatched under optimal conditions (6000 lux; 22°C), a process which takes approximately 72 hours. Once hatched, and with 24 hours of hatching, neonates were fed with an algae supplement, then transferred to test plates where they were separated into replicate wells containing the exposure substance at a range of concentrations. This kit allows for 5 different treatments plus a control, with 4 wells per treatment and 5 neonates in each well. Daphnia were incubated in the exposure medium and were maintained in the dark at 22°C; numbers of dead and immobilised organisms were assessed after 24 and 48 hours. In all cases, a range finding test was performed in the first instance, followed by at least two definitive tests. Results obtained for the three cytotoxins tested here showed the same rank order of potency as those reported in Zounkova (2007). That is, in this assay, cisplatin was the most potent compound. This is different than the order of potency observed in the algal tests where 5-FU is consistently observed to be the most potent of the three compounds. In both daphnia and algae, cyclophosphamide was found to be the least potent of the three compounds and in fact concentrations of several g/L were required to reach induce a response of 50% (Figure 6). The EC50 values (based on a mean of two repeat assays) of the three cytotoxins tested were as follows: CIS: 2.06 mg/L; 5FU: 243 mg/L; CYP: 2050 mg/L.
Figure 6: Effects of three cytotoxic drugs on immobilisation of Daphnia magna
Zebrafish assays
The zebrafish (Danio rerio) is an established model system in cancer research. It is used to study cancer development and progression and applied in screens to detect new anticancer drugs (Feitsma & Cuppen, 2008 ). Additionally the zebrafish is also commonly used in (eco)toxicology studies. Established OECD guidelines are available for an adult prolonged toxicity test (with a 2-week duration) (OECD 204); more recently, OECD test-guidelines were developed using zebrafish embryos for toxicity testing (OECD 236). As part of the Pharmas project we were investigating the effects of anti-cancer drugs on vertebrates. Cancer is defined as the uncontrolled fast growth of abnormal cells in the body. Most anti-cancer drugs disturb mechanisms of cell proliferation in order to stop the cancer cells from growing. During early development, normal cells are also proliferating very fast and disturbances during this period can have detrimental effects on development and pregnancy outcome. Therefore exposure to anti-cancer drugs during early development might be of higher risk. We investigated these potential risks by studying the effects of widely used anti-cancer drugs on the early embryonic development and on the larval stage of zebrafish and compared them with effects observed in adult zebrafish.
Adult zebrafish exposures
Adult zebrafish exposure studies were undertaken using a flow-through exposure system. Tanks of mixed-sex fish (6 males plus 6 females) were exposed to the test substance for 14 days according to OECD guidelines 204. A range of physiological end points were measured and the results assessed for all individuals and within the two sexes.
An initial pilot study was undertaken using three cytotoxic drugs, based on highest observed environmental concentrations, as well as the results obtained in the embryo toxicity assay undertaken by Partner 9. This pilot study used 10 µg/L cisplatin, 100 µg/L 5-FU and 100 µg/L cyclophosphamide. Following consideration of the results obtained in the pilot study, as well as the fact that cisplatin is an extremely expensive drug to use at high concentrations in a flow-through system, further studies looked in more detail at varying concentrations of 5-FU and cyclophosphamide.
Hepatosomatic index
In the pilot study, the hepatosomatic index (HSI) was observed to be significantly different to the controls when assessed using the student t-test (Figure 7). The livers of exposed fish were significantly smaller than those of the controls for all three drugs, with the most significant difference observed in those fish exposed to 5FU. In subsequent studies, the effect on HSI of exposed fish was not so clear; in one experiment using 5-FU (but where the fish were not as old as in the pilot study) there was no effect and in a subsequent study there was a dose-related trend in reduction of HSI, but the responses in individual treatments were not statistically significantly different from the control, as they had been in the pilot study.
Figure 7: Mean HSI of adult D.rerio exposed to either 100 µg/l 5FU, 100 µg/l CYP or 10 µg/l CIS for 14 days. A single asterix denotes significance at 5% in the students t-test, a double asterix significance at 1% when compared to the controls. Error bars show CI @ 95% Confidence Level.
qPCR
As the anti-cancer drugs are cytotoxic, it was suggested that this reduction in liver size could be the result of a loss of liver cells. In order to assess this, relative expression of RAD51 and p53, genes involved in the repair of DNA damage, in the livers of these animals were measured using quantitative rt-PCR. Primers were obtained from our partners in IVM. Two housekeeping genes were assessed, β-actin and elongation factor 1 α (ef1α). The results were normalised against the housekeeping genes using the Q-gene application, configured for “Mean Normalised Expression”, and tested for significance against the controls using the students t-test.
The data obtained using qPCR in these assays is not clear-cut. In the pilot study there appeared to be a consistent trend of the response of the genes analysed (that is, an increase in expression of RAD51 and a decrease in expression of p53) to all three chemicals tested. This pattern of response was not repeated in the concentration series experiments with 5-FU however (for full details please refer to D2.2). Currently the reasons for this are not clear, but it does seem apparent that there are few significant responses either at the morphological or genetic level in adult zebrafish which have been exposed to cytotoxic drugs, at relatively high concentrations, for a 2 week period.
A full description of the results obtained from these genotoxicity studies is given in Deliverable 2.2.
Zebrafish Embryo Toxicity
In general all tested anticancer drugs showed low toxicity in zebrafish embryos (LOEC at mM (g/l) range). The three compounds tested in more detail induced very specific effects. 5-FU: hyperpigmentation, p53 induction; CPA: short tail phenotype; Platins: none hatching. For CPA and 5-FU the results were similar to published data on rodent models and human studies. Cisplatin was the most toxic drug tested (LOEC 1 μM ). We could not detect effects on genes involved in DNA damage for any of the compounds at the concentrations tested. CPA and Cisplatin are known to bind to DNA and crosslink it. Despite the similar mechanism they induce very different effects in zebrafish embryos and larva. Therefore we conclude that endpoints of DNA damage are not the most sensitive ones for anticancer drug exposures. CPA, 5FU and Cisplatin showed differences in the developmental windows when the embryo is the most sensitive to the drugs. Cisplatin and CPA are more toxic during the early phase of the development whereas 5-FU shows effects at the later stages.
Furthermore, we could not find any literature comparing effects of anticancer drugs at different developmental stages of zebrafish.
For a more comprehensive description of the results obtained from the work undertaken for this workpackage see deliverable report D2.2.
Cisplatin and Carboplatin
These compounds both affected hatching of zebrafish embryos. It is known that platinum substrates can bind to metalloproteases and inhibit their function (Arnesano et al., 2009) . The hatching enzymes of zebrafish are metalloproteases. We could not detect any effects on gene level or visual malformations that might indicate that the excretion of the hatching enzymes is affected. Nevertheless we could clearly see that the chorion of exposed embryos is not digested and therefore prevents hatching. Other studies also detected a postponed hatching when embryos are exposed to platinum complexes like PtCl2 (Osterauer et al, 2011 ). We assume that Cisplatin and Carboplatin and other platinum complexes bind to the hatching enzymes and inhibit their function. The hatching process is highly conserved between different fish species (Sano et al., 2008 ). Therefore the effect of platinum complexes could be a general effect for most fish species.
Cyclophosphamide(CPA)
CPA induced very specific effects in exposed zebrafish embryos. The most sensitive time for exposure was after organogenesis has started. This fits with reports on other species. Also the effects on body size were comparable. The decreased heartbeat could be detected at concentrations where the total length was already reduced. Therefore the size reduction is not caused by a reduced heartbeat. No genes linked to DNA damage were induced. We assume that CPA induces oxidative stress followed by reactive oxygen production and increased oxygen consumption which leads to the observed effects (body size, eye size, heart beat) and cell death (gene induction).
5-Fluorouracil
5-FU was tested at concentrations up to 20mM (near the solubility threshold of 5-FU in water). No teratogenic effects could be detected besides slight hyperpigmentation at levels higher than 10mM. No genes linked to DNA damage were induced. Only p53 was elevated in exposures after 3dpf at 10mM concentrations. P53 induction can lead to hyperpigmentation and might explain what we could observe in exposed embryos. Our findings show that 5-Fu is up taken and can be converted into its active form in zebrafish embryos older the 3dpf. Why we could not detect effects at lower levels and before 3dpf (TS protein increase, 5-FU degradation, low uptake) is still unclear.
Comparison of sensitivities of adult vs embryo zebrafish stages
From the studies carried out with embryonic, larval and adult stages of zebrafish in Pharmas, compound- and life-stage specific sensitivity to the anti-cancer drugs was observed, albeit at concentrations higher than those found in the environment. Cisplatin showed a life-stage specific effect on hatching during early development at the lowest effect concentration of the three compounds tested in embryos (0.001 mM or about 300 µg/l). In adults, 10 µg/l cisplatin did not show any effects on the endpoints measured. Higher concentrations were not tested, making it difficult to conclude if adult or early life stages are more sensitive. For 5-FU, effects on HSI were found in adult fish at lower concentrations than those observed in embryos and larvae, suggesting higher susceptibility of adults to this compound, likely due to the enhanced bioactivation of 5-FU by the adult liver or lack of uptake in embryos. For CPA, a life-stage specific effect on growth during early larval development was observed, though at concentrations higher than those tested in adults. CPA at 100 µg/l induced a significant reduction in HSI, indicating an enhanced sensitivity of adult stages to this drug.
The effects of anti-cancer drugs observed in Pharmas were at concentrations far exceeding environmental concentrations. It should be mentioned, however, that the exposure duration used in this study was rather short, especially those used for embryotoxicity testing, and we cannot rule out that longer term exposures may result in effects at lower concentrations. Indeed, complete life cycle studies with 5-FU in zebrafish revealed genotoxic effects at much lower concentrations than used here (Cytothreat presentation, Brussels 2013). We also focussed on phenotypical effects of anti-cancer drugs on zebrafish development, and have not assessed the long term effects of early-life stage exposure on reproduction, growth or survival.
C.3. Mixture substances
The toxicology and ecotoxicology of chemical mixtures has one strong, common pattern – independent of the exposed organism, analyzed biological endpoint and chemical composition of the mixture: the joint toxicity of a combination of chemicals is in the overwhelming majority of cases higher than the toxic effects of each single compound. This implies that chemical risk assessments and guidelines need to take combination effects into account – or otherwise run the risk of a serious underestimation of the actual risk for humans and the environment.
The classical toxicological concepts of Concentration According (CA) and Independent Action (IA) have been successfully used in a range of scientific studies to predict the (eco)toxicity of various chemical mixtures, including combinations of pharmaceuticals. CA is thought to be applicable to mixtures of similarly acting compounds, while IA is based on the idea that the components in a mixture act independently, via different physiological pathways on a common biological endpoint. CA and IA make use of information on single substance toxicities for predicting the joint action of a mixture.
Consequently, both concepts assume that the environmental or human health risk of a mixture is dependent on the number of present compounds, their individual toxicities and their concentrations in the mixture. Although several antibiotics and anti-cancer drugs have been detected in aquatic systems, neither have the available analytical fingerprints been analyzed from a mixture toxicity perspective, nor has the expected exposure to pharmaceutical mixtures been modeled on a broader scale and then analyzed for its potential risk.
In this context, works have been performed in order to (i) assess the environmental risks of realistic mixtures of antibiotics and/or anti-cancer drugs; (ii) analyze by how much current regulatory single-substance oriented risk assessment strategies run the risk of underestimating the actual risk of environmentally realistic exposure situations; and (iii) provide the necessary scientific tools and regulatory options for considering the joint effects of antibiotics and anti-cancer drugs during the environmental risk assessment process. To reach these objectives, the assessment of realistic mixtures of antibiotics and anti-cancer was carried out on fish (embryos) and selected species of bacteria and algae, on the biodiversity and ecological succession in natural microbial communities, on bacterial communities in STP plants and by the optimisation and validation of predictive mixture hazard modelling approaches
The prediction of risk quotient for environmentally realistic mixtures of pharmaceuticals was performed in several steps. A review of the scientific literature on monitoring of mixtures of pharmaceuticals was conducted firstly to lead to a selection of one particular publication with the broadest range of investigated pharmaceuticals, in particular it is one of the very few published studies where antibiotics were included at least to a certain extent (7 of the 26 studied pharmaceuticals were antibiotics). The ecotoxicological data for all the 26 monitored pharmaceuticals were collected from the peer reviewed literature and publically accessible databases. Data on toxicity and ecotoxicity for mammals, birds, higher plants, fungi and bacteria were only available for a very few compounds and organism groups. Those available were grouped in three groups: algae, invertebrates and fish. Finally, the measured environmental concentrations (MECs) and the compiled ecotoxicological data were used for calculating the toxic units (TU = MEC/EC50) of the individual pharmaceuticals for the three organism groups, and the resulting risk quotients (as sums of toxic units, STU) of the mixtures:
All analyses had to be confined to the summation of TUs (i.e. the mixture toxicity assessment according to CA), as the available data did not allow the calculation of the corresponding IA predictions. We therefore also estimated the maximum possible error that might result from this simplification.
A total of 890 individual ecotoxicity values were collected. The STU for algae and invertebrates is consistently higher than 10-3. This implies that the exposure situation might give reason for concern, if the standard single substance assessment factor of 1 000 for the base set of data (REACH guideline) is applied. In fact, mixture risk quotients are then between 16 (Swedish STP effluent) and 45 (French STP). Whether there is an actual risk for the aquatic environment as a consequence of STP discharge then ultimately depends on the actual dilution of the STP effluent in the receiving river.
It might be argued that the application of CA is potentially overconservative. The strict mathematical relation between CA and the competing concept of IA allows to calculate the maximum possible overestimation by CA (in relation to IA); the factor between both predictions being usually between 1.5 and 3, only in two cases (STU for fish for S1-F and L1-F) the ratio is between 4 and 5.
In view of all other uncertainties inherent in the exposure as well as the effect data, a mis-prediction by a factor of 1.5 to 3 might be regarded as of minor concern. In fact, the ratio IA to CA predicted EC50 is well below 2 for all STPs if algae (the most sensitive organism group) are considered. That is, even if a mixture assessment would be based on IA, the joint toxicity of the pharmaceuticals might still be of concern.
The sum of PEC/PNEC ratios (or MEC/PNEC ratios) can be considered a pragmatic approximation of CA, while a STU based risk assessment is conceptually closer to the concept of CA. A comparison between the risk quotient that results from summing up MEC/PNEC ratios and the risk quotient that results from STUs showed that the resulting ratio is less than 1.3 in all cases, indicating that summing up MEC/PNEC ratios (respectively PEC/PNEC ratios) can indeed be used in a first tier risk assessment of pharmaceutical mixtures. The toxic unit distributions (Figure 8) are specific for each organism group and effluent. However, all distributions have a common shape, which is characterized by the fact that a comparatively few compounds always contribute the bulk of the toxicity. In fact, the first 10 compounds always provide more than 99% of the total STU. The detail of each distribution for all STP tested are described in deliverable 3.1.
Figure 8: Example of Toxic Unit Distribution (relative) for algae and invertebrates for a STP‐effluents
Deliverable .3.3 focused then on providing tools to allow also new users to implement Concentration Addition for predicting and assessment mixture effects. These Mixture Tools provide three sets of tools and analysis instruments for prediction and assessment of mixture toxicities. The first toolset provides a framework for the calculation and visualization of mixture toxicities, based on the assumption that sufficient input information is available for the calculation of the toxic units of the individual compounds. This toolset also provides an overview of the quantitative consequences of synergistic interactions. The second toolset comprises a set of functions that can be used to calculate the predicted mixture toxicity according to Concentration Addition and Independent Action in arbitrary scenarios. The third toolset provides a set of concentration-response functions for estimating ECx values (e.g. EC50's, EC10's), which then can be used for the mixture toxicity assessment. Examples for all toolsets are described in deliverable 3.3 and are accompanying the Mixture Tools collection.
Empirical analyses using a range of mainly bacterial assays were conducted in the context of deliverable 3.2. A mixture of 14 pharmaceuticals detected in the effluent of the Ryverket STP in Gothenburg, which includes 7 antibiotics, was tested for its toxicity towards biofilm communities (periphyton), sampled from a pristine site. The reconstituted mixture did not provoke significant effects on bacterial communities at the effluent concentration. The total mixture causes highly significant effects only at a concentration of 100x times the effluent, while first effects of the antibiotic sub-mixture become visible at roughly 10 times the effluent concentration (EC10=10.3x the effluent concentration). Applying an assessment factor of 10 - which is deemed appropriate, given the fact that the available background data from single species assays indicate that bacteria are indeed the most sensitive trophic level and higher-tier tests with fish or invertebrates might hence not provide substantial new risk-relevant information - results in a final empirical risk quotient for the pharmaceutical part of the effluent of almost exactly one. This fits well with the risk estimate of 20 that was put forward in deliverable 3.1 based on predictive modeling and used an assessment factor of 1 000. The results hence (i) strongly support the current regulatory strategy to base the environmental assessment of antibiotics - and consequently antibiotic-dominated mixtures - on tests with bluegreen algae (EMA, 2006), (ii) emphasizes that local conditions, in particular the dilution factor of the receiving waters, need to be taken into consideration for a realistic risk estimate.
For the work with STP-derived bacterial communities derived from STP effluents the mixture of pharmaceuticals were tested in different concentrations. Different community-level toxicity endpoints like cell viability and multiplication, respiratory activity and metabolic diversity were addressed. Also single species toxicity testing on Vibrio fischeri and Pseudomonas putida were performed. Contrary to V. fischeri, STP derived bacterial communities were not affected by pharmaceutical mixtures in the range of measured effluent concentrations. This could be an indication that these communities are already adapted to mixtures of pharmaceuticals in actual measured concentrations. Details on the experimental work are given in deliverable 3.2.
Tolerance development was further investigated in the work on deliverable 3.5. As a worst case scenario, we investigate the impact of treated effluent from a plant receiving waste water from bulk drug production in Patancheru, India and its main antibiotic ingredient, ciprofloxacin, to natural bacterial communities (periphyton biofilms). Toxic effects were assessed as changes in carbon utilization pattern using Biolog Ecoplates™. Bacteria exposed in Ecoplates directly after sampling from an uncontaminated site were highly sensitive towards both ciprofloxacin and effluent. Observed EC50 values for the average ability to degrade carbon source were 70.3 nmol/L for pure ciprofloxacin , respectively 0.08% for the effluent. The degradation of specific individual carbon sources was even affected already at concentrations a tenfold lower. These concentrations are lower than the effect concentrations observed in previous studies, indicating that natural periphyton might provide suitable test material for investigating environmental effects of antibiotics. The high toxicity of the effluent also clearly indicates the need for appropriate risk assessment and management strategies for the investigated effluent.
Pre-exposure to 0.1% of the effluent clearly induced tolerances both to the effluent itself, but even more so to pure ciprofloxacin. The EC50 of ciprofloxacin was increased by a factor of al-most 4. This substantial development of tolerance against ciprofloxacin caused by the effluent indicates the possibility of resistance development in the environment that might potentially hamper the use of fluoroquinolones in human medicine.
C.4. Transformation products
After usage, pharmaceutical products will undergo transformations starting from human metabolism to degradation in conventional and advanced effluent treatment, in environmental processes and finally during drinking water treatment. The majority of pharmaceuticals taken by a patient will be at least partially excreted in the form of metabolites or conjugates. Often degradation in sewage and water treatment and the environment is incomplete resulting in stable transformation products. The formation and presence of such stable transformation products in the effluent of sewage works, surface water, and drinking water treatment is reported in the scientific literature with increasing frequency (see for example Kosjek and Heath 2008 , Perez and Barcelo 2007 , Gröning et al. 2007 , Vasconcelos et al. 2009 , Längin et al. 2009 ).
This is even more of importance as advanced oxidation techniques employing e.g. ozone, hydrogen peroxide, light or electro-coagulation are increasingly under discussion for the removal of pharmaceuticals and other micro pollutants in effluent treatment and drinking water treatment. In general the literature dealing with TPs is “exploding”. Because of non-standardized approaches, the number of possible treatment technologies and individual test conditions it is difficult to keep an overview on all work published for pharmaceuticals in general.
Toxic compounds may be generated within treatment formed. In some cases the transformation products can be even more toxic, e.g. genotoxic, than the parent compound (e.g. Isidori et al. 2005 , 2007; Lee et al. 2007 ; Wei-Hsiang and Young, 2008 ). Through the application of analytical techniques for structure elucidation, the chemical structure of such compounds can be established. Unfortunately, for most transformation products only little information on their properties is available for two reasons: 1) They are not available in amounts necessary for experimental testing as isolation and/or targeted synthesis is too expensive and 2) the number of compounds described in literature is too high to be tested for reasons of time and cost. Therefore, data for a full and reliable risk assessment of parent compounds and their resulting transformation products is missing.
A solution for this undesired situation is offered by the employment of computer based methods (“in-silico screening and assessment”, for an overview see Ekins 2007 ). Using in silico screening tools only the structural formula is necessary to identify the compounds of highest interest with regard to their physico-chemical and toxicological properties.
In this context, the project aimed
• to collect knowledge on stable transformation products of antibiotic and antineoplastic compounds (parent compounds selected within in the PHARMAS project) and assess experimental approaches by an extensive literature review,
• to study experimentally the fate of parent compounds selected within in the PHARMAS project and TPs generated different processes proposed for water and waste water treatment and environmental processes (effectiveness and treatment times, identification of TPs and assessment of the toxicity of TPs using experimental and in-silico methods),
• and to develop a first basis for the targeted design of pharmaceuticals with enhanced degradability and elimination behaviour (“benign by design”).
An extensive literature search was conducted in the scientific databases SciFinder (database of CAS), Web of Knowledge, and PubMed. The aim of this part of work was to investigate whether TPs of the target compounds selected within the PHARMAS project are formed during the different types of treatment processes for effluents and water and to collect information on their chemical structure, respectively. The treatment processes, which might lead to transformation products, were: photolysis, biodegradation, chlorination, ozonation and other advanced oxidation processes. Results of additional studies, in which more complex treatment processes were investigated, were also taken into account, e.g. formation of TPs during combinations of the above mentioned ones (e.g. chlorination and ozonation) ozonation combined with biological reactors or degradation studies in outdoor microcosms which simulated complex natural water environment.
The literature study shows that the knowledge on transformation products formed is scarce despite a steep increase of publications dealing with this issue in recent years. However they focus most often on a narrow spectrum of compounds using different treatment conditions (details often not sufficiently detailed provided). Out of all selected pharmaceuticals of PHARMAS for four, namely imatinib, capecitabine, hydroxycarbamide and ifosfamide, there was completely no information available. Interestingly, this is a mixture of old and new drugs. For others, such as ciprofloxacin much more than fifty different TPs were described in literature relating in some cases to different conditions of there formation. Studies focusing on the transformation products of pharmaceuticals formed by chlorination are currently only few for all compounds. No reports were found which have dealt with this topic for selected cytostatic drugs. In general there is more information available on TPs formed from antibiotics than from cytostatic drugs. The largest part of older studies focused on the investigation of the elimination kinetics of parent compounds without presenting any information on the chemical structure of individual TPs formed and their amount formed. In more recent publications there is an increasing trend to deliver such information.
In a minority of the publications dealing with the compounds that were of interest within the Pharmas project there was information presented which was related to the identification of TPs formed and/or on the assessment of the biological potency of them when present in mixtures with parent compound. In most of the cases TPs were “identified” by using most often mass-spectrometry analysers such as LC-MS/MS, MSn, a few using more advanced high resolution mass spectrometry. Based on the MS results chemical structures were suggested. The reliability of such proposed structures of TPs could not be assessed in all cases on the basis of the information presented. On the other hand there were also studies that focused on appearance of TPs based on comparison with analytical reference standards if available at all (e.g. human metabolites). So far there were only few lab based studies on the further fate or effects of single TPs. In most of the cases, studies focussed on structure elucidation and pathways of TPs formation. Some of the studies assessed additionally the toxicity of mixtures of TPs formed after a certain time of treatment or biodegradability of this kind of mixtures. Due to cytotoxic and genotoxic behaviour of cytostatic drugs an impact on the environment could be expected but was not assessed. There is usually a decrease of antibiotic activity among TPs formed from antibiotics. However there are examples of increased toxicity of those TPs or no significant difference in toxicity between TPs and parent compounds. The details of the review are collected in the deliverable 5.1.
In order to enhance quality of drinking water and to eliminate pollutants in sewage water, several techniques have been employed for many years. With emerging pollutants, new techniques have come into focus. Treatment steps for both kind of water include generally (depending of the size of the plants) retention (filtration, flocculation, sedimentation) coupled to biodegradation and advanced processes (oxidation by different chemicals, disinfection, photolysis. Retention processes consist of the transfer of the substances of interest from the liquid to a solid phase and the elimination of this last one. Contrarily, degradation processes conducted either result in the total elimination of the substances (mineralization if no sorption) or to the incomplete transformation resulting in products that need to be characterized (in particular in terms of toxicity) or are totally inefficient. The outcome of such treatments depends heavily on the compound and the type of the applied treatment. A number of active ingredients of pharmaceuticals have been found to be persistent in regard of wastewater treatment as well as drinking water treatment. Thus, although treatment of drinking water and sewage water is well established, there are still many treatment-resistant substances present. Thus the improvement of the efficacy of the treatment processes leads to the formation of transformation products that can pose a risk for the environment and the public health.
Based on the above mentioned results, the chemical behaviours of a selection of antibiotics and cytostatic drugs in selected sewage and drinking water treatment processes were investigated obeying the following scheme (Figure 9).
Figure 9: General approach for the investigation of the fate and biological effects of TPs the selected pharmaceuticals
For the further investigation, a selection of pharmaceuticals of both classes was done based on literature review. According to a recent review, 5-fluorouracil, cyclophosphamide and methotrexate were the most widely consumed anticancer drugs worldwide. Imatinib belongs in contrast to the above mentioned group of newer anticancer drugs on which nearly no information was available. Ciprofloxacin is a heavily used antibiotic that was shown not to be biodegradable in sewage treatment in several studies. The cytotoxics methotrexate, cyclophosphamide, imatinib and 5-fluorouracile, and the antibiotic ciprofloxacin were therefore investigated in laboratory and pilot scale studies. Slightly different approaches were developed for treatment of drinking water and sewage/sewage plant effluent (see Error! Reference source not found.9) to address the differences related to drinking water on the one hand and waste water on the other one. In total, for the five pharmaceuticals investigated in this study, eight different treatments from drinking water treatment and sewage water treatment were tested. All pharmaceuticals gave transformation products in at least in one treatment; for most of these, no finally reliable structure could be proposed. The following treatments were investigated: LI (low intensity) and HI (high intensity) UV-photolysis, Xenon photolysis, chlorination, ozonation, and a combination of the latter two.
5-FU, imatinib and related cytotoxcis were determined in the water samples. As we could only find TPs for the treated imatinib and CPA the samples were only run with those settings (positive ionization and PFP and HILIC LC columns respectively). Considering the low levels of IB found in the water samples (<5 ng/L) it is not likely to detect the TP estimated to constitute ca 5% of the formed degradation products in the water samples, assuming a similar LOD (5ng/mL). Also, neither imatinib related TPs nor CPA TPs could be detected. A clean up method targeting the TPs specifically would be necessary to be able to determine their presence and concentration.
5-FU was fully degraded by chlorination and ozonation. After both treatments the formation of TPs has occurred (as evidenced by DOC monitoring), whereas chlorination after ozonation degrades the formed TPs. 5-FU was not affected by xenon lamp or biodegraded, but UV fully degraded 5-FU. TPs were also expected to be seen after UV lamp treatment. Probably due to the small molecular size of this anti-cancer drug, and the polarity of the fractions, no masses could be detected and isolated in with the analytical detect, neither in positive nor in negative ionization mode. This example demonstrates that in some cases for TP identification other strategies need to be considered.
CPA was not affected by chlorination or Xenon lamp treatment, which indicates that there is no full mineralization of the compound. Hence, TPs could be expected (confirmed also by DOC measurements). It was degraded (ca. 75%) by ozonation, UV lamp and biodegradation. In other words, none of the treatments fully removed CPA. Two CPA TPs were found, one after ozonation and one after biodegradation.
Imatinib was fully degraded with chlorine and ozonation, partially by UV lamp treatment but not by biodegradation. The DOC measured after chlorination, ozonation and UV lamp treatment indicated TP formations.
Regarding ciprofloxacin and methotrexate, structures for transformation products resulting from chlorination and UV-photolysis could be proposed.
Parent compounds and the mixtures resulting from LI (low intensity) and HI (high intensity) UV-photolysis, Xenon photolysis, chlorination, ozonation, and combination of the latter two were tested for their toxicity with the AMES Aqua test. Toxic effects were not observed in any of these scenarios. Even the parent compounds (with the exception of ciprofloxacin) were not toxic at the concentrations applied. The tested concentrations did not pose an acute toxic potential within the tested parameters of the AMES aqua test.
TPs were additionally tested with quantitative structure activity relationships (QSAR) regarding various additional toxicological endpoints if an unambiguous chemical structure was available. Regarding methotrexate, chlorination did not result in less toxic transformation product. One transformation product formed during LIUV-photolysis had even more alerts as the parent compound thereby giving hints of increased toxicity through treatment processes. Proposed transformation products for ciprofloxacin possess mostly the same alerts as the parent compound. Here, no change in the pattern of the toxicity could be found. One transformation product (chlorination M 2_2) gave much less alerts as the other.
Summarizing treatment processes, all investigated pharmaceuticals formed at least during one treatment transformation products. Most of them are not known. Further work is needed to confirm the identity of the TPs. Drinking water treatment processes including chlorination lead preferably to transformation products that contained an atom of chlorine which is not first hand to be expected within photolysis treatment (corresponding to sewage water treatment processes). A reason could be that by photolysis active chlorine containing species are generated from inorganic chloride. As most of the transformation products could neither be identified nor fully assessed regarding toxicity, further research is needed. In some cases structures could be put forward but distinction of different isomers was not possible and thereby preventing QSAR analysis.
The results have been collected in deliverable 5.2.
By applying computer-based methods such as QSAR it could be demonstrated that relevant alerts for non-degradability could be identified and neutralized in order to construct a sustainable version of the parent compound. This strategy was applied to ciprofloxacin and its transformation products, but evidence was found that it is most likely also applicable to other substances. Furthermore, molecular docking software could help to provide a first indication that the altered compound might still be pharmacologically active without the need to synthesize the molecule beforehand and subsequently to test it in the wet lab (deliverable 5.3). Candidates which pass this in-silico validation procedure can be selected for further investigations and might possess opportunities to generate alternative or improved pharmaceuticals. In this respect the applicability and advantage of in-silico tools as a step in the direction to benign by design pharmaceuticals of the future was shown. Some first guiding principles were identified: With ciprofloxacin as an example for constructing environmentally improved pharmaceuticals by computer methods (so-called “benign by design” process) was specifically demonstrated. This approach will pave the way to developing practicable ways to design degradable pharmaceuticals (see deliverable 5.4).
C.5. Risk assessment
A specific task was dedicated to risk assessment with the aim to integrate the information concerning exposure, effect of single and mixture products in order to estimate the environmental risk posed by the use of the selected pharmaceuticals for a selected number of endpoints. More specifically, the main objectives were:
1. Set up realistic worst case exposure scenarios for exposure of sensitive subgroups of the human population to a selected number of pharmaceuticals;
2. Characterize the cumulative risk due to exposure of humans and non human endpoints (aquatic ecosystems, and wildlife) to a selected number of pharmaceuticals;
3. Assess the uncertainty and variability involved in the risk characterization by means of probabilistic techniques and expert assessment of uncertainty;
4. Explore alternative risk indicators which are more informative than the conventional PEC/PNEC ratio.
To reach these objectives, a screening tool was developed to identify combinations of human pharmaceuticals, locations and exposure groups in Europe for prioritization (deliverable 4.1). The screening tool was adapted into a location-specific decision support tool to be used by physicians for the prescription of drugs. Furthermore, an extensive uncertainty analysis was performed. At first, this analysis focused on quantifying uncertainty in the acceptable daily intakes of two pharmaceuticals, i.e. ciprofloxacin and methotrexate. Subsequently, uncertainty in human and ecological exposure estimates was quantified by means of Monte Carlo simulation. An expert workshop was organized to assess the sources of uncertainty that cannot be quantified in a Monte Carlo analysis. Finally, a study was performed to optimize the use of information in ecological risk assessment of pharmaceuticals using the multi substance Potentially Affected Fraction (msPAF) as an endpoint.
The main outputs are resumed below.
A European wide, spatially explicit screening tool was developed that was used to identify locations in Europe showing the highest human and ecological risks due to the use and presence of 11 selected human antibiotics and 7 antineoplastics.
For human risk assessment, a range of exposure scenarios was defined, i.e. based on age (0-1 years, 1-10 years, 18-65 years and >65 years), drinking water purification technique (none, conventional and advanced), diving behaviour (none, recreational and professional; 18-65 years only), and food origin (local and Europe-wide). Human contact media that were taken into account were drinking water, fruits and vegetables, meat products, milk products, fish, surface water and soil. HD50 and HC50 values were used to characterize the human and aquatic toxicities of the pharmaceuticals, respectively. These toxicity values represent the hazardous dose (HD50) or hazardous concentration (HC50) of a chemical at which at least half of the individuals in half of the species is affected. HD50 and HC50 values were based on acute toxicity data, due to data availability and under the assumption that relative acute and chronic toxicities of the substances are similar. The risk quotients for all combinations of substance, location and exposure group were calculated by division of concentrations (for the aquatic environment) or daily intakes (for human health) by the toxicity values. Finally, summed risk quotients for all antibiotics and for all antineoplastics were calculated per grid and exposure group (for human health) on the basis of the concentration addition principle. Figure 10 shows that the summed aquatic risk quotients due to exposure to antibiotics are substantially larger than those due to exposure to antineoplastics, a pattern that is also visible at the level of the individual compounds.
Figure 10: Summed aquatic risk quotients (predicted surface water concentrations / HC50) due to exposure to (A) antibiotics and (B) antineoplastics. The grids with the highest summed risk quotients are outlined.
The highest aquatic risk quotients are calculated for three antibiotics: levofloxacin, doxycycline and ciprofloxacin. This can be explained from their aquatic toxicities (i.e. second, third and fourth most toxic of all compounds assessed), as well as their very low removal in both STP and the environment. Risk quotients were highest in the Milan region in northern Italy. This can be explained from the high levofloxacin consumption in Italy, compared with the other EU Member States, and the high population density of the region.
Highest human health risk quotients are estimated for infants that consume conventionally treated drinking water. Figure 11 shows that the difference between the summed health risk quotients due to exposure to antibiotics and due to exposure to antineoplastics is not as clear as for the aquatic environment (Figure 10).
Figure 11: Summed health risk quotients (predicted exposure / HD50) for infants due to exposure to antibiotics and antineoplastics, after consumption of conventionally treated drinking water and either locally produced or Europe-wide food. A = antibiotics and locally produced food; B = antineoplastics and locally produced food. The grids with the highest summed risk quotients are outlined.
There is a large influence of food origin on the health risk quotients. 5-Fluorouracil and to a lesser extent ciprofloxacin have the highest health risk quotients when a local food origin is considered. In the two grids located in eastern Spain (Figure 11), sludge application on agricultural soils is very high and run-off from agricultural soils is expected to be rather low.
The tool is to our knowledge the first that can be used to prioritize combinations of human pharmaceuticals, locations and exposure groups, in such a way that potential hot-spots and substances can be identified for future monitoring activities and future research in the framework of aquatic and human health risk assessment. All the details have been reported in deliverable 4.1 and have been published in an international scientific journal
In a second task, we explored alternative endpoints for human and ecological risk assessment. In order to realize this aim, the screening tool (described above) was adapted to make it suitable for the assessment and comparison of the environmental impact of two alternative pharmaceutical prescriptions. This methodology provides physicians with the opportunity to include environmental considerations in their choice of prescription. A case study with the two antibiotics ciprofloxacin and levofloxacin at three locations throughout Europe showed that the preference for a pharmaceutical might show spatial variation, i.e. comparison of two pharmaceuticals might yield different results when prescribed at different locations (Figure 12). This holds when the comparison is based on both the impact on the aquatic environment and the impact on human health. The relative impacts of ciprofloxacin and levofloxacin on human health were largely determined by the local handling of secondary sludge, agricultural disposal practices, the extent of secondary sewage treatment, and local food consumption patterns. The relative impacts of ciprofloxacin and levofloxacin on the aquatic environment were mostly explained by the presence of specific sewage treatment techniques, as effluents from sewage treatment plants (STPs) are the most relevant emission pathway for the aquatic environment.
Figure 12: Relative impact of ciprofloxacin and levofloxacin on the aquatic environment (Aq.) and human health (Hum.), in northern Italy (left), south-eastern Spain (middle) and eastern Sweden (right). Blue bars: relative impact of the prescription of one daily defined dose (DDD) of ciprofloxacin; Yellow bars: relative impact of the prescription of one DDD of levofloxacin.
The use of alternative risk assessment endpoints was furthermore addressed in a study that focused on ecological risk assessment of pharmaceuticals. The main aim of this study was to optimize the use of available ecotoxicological data to assess the risk of multiple pharmaceuticals to aquatic ecosystems. In this study, the use of the multisubstance Potentially Affected Fraction (msPAF) was used as an alternative endpoint to quantify the impact of multiple substance on aquatic ecosystems. A combination of concentration addition, response addition and toxic pressure addition was used to predict the combined impact of multiple substances. Bayesian techniques were used to combine substance-specific ecotoxicological data with other sources of information contained in ecotoxicological databases, thereby reducing the uncertainty in the msPAF. The analysis was performed for a realistic mixture of pharmaceuticals in the German Ruhr area, as predicted by the prioritisation tool outlined in D4.1. The study is described in detail in Deliverable 4.4.
Finally, a detailed and integrated probabilistic risk assessment for a selected number of high risk exposure scenarios proposed. Several studies and activities were performed to realize this aim. First, the screening tool for the location-specific prioritization of human pharmaceuticals in Europe was subjected to a probabilistic assessment with Monte Carlo techniques. This way, the uncertainty underlying the calculations was quantified. Uncertainties were analysed in an integrated way; combining ecological and human risk assessment where possible. Next to a complete assessment of the whole chain of calculations from consumption data to risk quotients, the tool was divided into six sub models, which were then separately assessed. These six sub models were 1) Emissions to wastewater, 2) STP removal, 3) Environmental fate, 4) Drinking water purification, 5) Human exposure, and 6) Toxicity and effect. Three substances were selected for a case study, being ciprofloxacin, erythromycin and 5-fluorouracil. They were chosen because they represent both antibiotic (ciprofloxacin and erythromycin) and antineoplastic substances (5-fluorouracil), and because they differ in their ionizing properties, i.e. ciprofloxacin is a zwitterionic compound, erythromycin is a base, and 5-fluorouracil is an acid. Furthermore, three STPs in three different grids were selected for the case study. This selection was based on the range in treatment techniques applied between the three STPs, and the variation in the extent to which secondary sludge is disposed of between the three Member States of concern. Parameters were assigned a (log-)normal distribution, preferably based on literature data or, if data were insufficient, based on related data. An exception was made for parameters that represent fractions, which were characterized with a beta-distribution, based on literature data. Parameters for which a quantification of the uncertainty was not possible on the basis of literature data were assigned an uncertainty distribution based on expert judgement. The results of this study are described in detail in Deliverable 4.2.
Second, a study was performed to separately quantify different sources of uncertainty and variability in the derivation of the Acceptable Daily Intake (ADI) of two pharmaceuticals, i.e. ciprofloxacin and methotrexate. A separate propagation of all sources of true uncertainty and interindividual variability throughout the derivation process enables policy makers to set separate standards for the fraction of the population to be protected and the confidence level of the exposure limit protecting that fraction. Additionally, the methodology enables the inclusion of potentially sensitive subpopulations, and it identifies the main sources of uncertainty and variability influencing the exposure limit, therefore providing a focus for future research. For both substances studied, the ADI was mainly influenced by uncertainty in the derivation of a point of departure, the extrapolation from sub-chronic to chronic toxicity, and by uncertainty in the extrapolation from mammalian to human toxicity. It was shown that traditionally derived ADIs protect 95%-99% of the population with 95%-99% confidence. The framework for deriving probabilistic ADIs is illustrated in Figure 13. The results of this study are described in detail in Deliverable 4.3.
Figure 13: The framework for the probabilistic derivation of ADIs (U = uncertainty; IV = interindividual variability; S = systematic difference).
Third, an expert workshop entitled Uncertainties in Human and Ecological Risk Assessment of Human Pharmaceuticals was organized on 24-25 October 2013 in Ravenstein, The Netherlands. The main aim of the workshop was to identify, describe and assess the uncertainties currently involved in risk assessment of human pharmaceuticals. This information is particularly important for planning future research to further optimize the risk assessment process. The 26 experts that participated in the workshop were divided over 6 working groups, i.e. (1) emissions to wastewater, (2) STP removal, (3) environmental fate, (4) drinking water purification, (5) human exposure, and (6) toxicity and effects. Each working group identified, described and assessed the uncertainties within their particular theme, using the screening tool as a starting point for the discussions. The assessment and prioritization of the uncertainties proceeded according to an established methodology which was supervised by an external expert, i.e. dr. Anne Knol. The workshop provided detailed insight in the main sources of uncertainty involved in assessing the human and ecological risks of human pharmaceuticals. A detailed analysis of the results is ongoing and will be processed into a joint scientific publication of the workshop participants that is expected by the end of 2014.
Potential Impact:
A European risk and hazard classification system for medicinal products would be a strong core element of precautionary EU policies for minimising possible risks for human health and ecosystems due to environmental exposure to pharmaceuticals and their derivatives. Such a system would primarily be designed as a (web-based) decision support tool for practitioners (e.g. physicians and pharmacists) providing straightforward action alternatives. There is initial evidence from a five year experience with such a system in Sweden that it can influence prescription, sales and consumption patterns of pharmaceuticals in terms of promoting less problematic substances where therapeutically equivalent alternatives are available (LIF, 2007). Based on the Swedish experience, there have been on-going discussions on the possibility of extending t his scheme to the rest of the EU. At the same time, concerns raised in the completed EU-funded project KNAPPE indicated that the lack of available data on the effectiveness of the scheme could be a barrier to its implementation in other countries. Some stakeholders see a thorough evaluation of the experiences with the Swedish “pilot” system as a prerequisite for its extension. The stakeholder community for such a system is most certainly a wider group than just practitioners and pharmacists; data on pharmaceuticals may also serve as aid in decision-making for drinking water plant managers, river basin managers, managers of waste water treatment plants etc.
In this context, and in addition to the scientific and technical works described above, another task more focussed on stakeholders and society was performed. The main aims were
- to explore stakeholder requirements on an EU wide system, to assess its influence on prescription, sales and consumption patterns of selected pharmaceuticals, to estimate its socio-economic impact, and, based on these findings, to develop a prototype tool that in particular accounts for the different markets, health care systems and cultural background of EU member states.
- to identify stakeholders' level of information and understanding of the issue, as well as associated risk conceptions; this will help define the characteristics required of possible risk communication alternatives, as well as influence project dissemination.
To answer these objectives, three main tasks have been implemented: (i) a qualitative socio-empirical study on the specific requirements of the European stakeholder community regarding such a classification system; (ii) a Delphi-based socio-economic impact assessment of an European classification system using the example of three countries and comparison to the Swedish system; (iii) a prototype database for risk and hazard classification of pharmaceuticals.
For the qualitative socio-empirical study on the requirements of different stakeholders on a risk and hazard classification system for pharmaceuticals in the environment, a short inventory of the specific European stakeholder community was set up as a first step and the perspective of stakeholders on the issue was explored via in-depth phone interviews with a group of key stakeholders. The questions focused on:
• the stakeholders’ attitude towards such a possible system,
• their evaluation of its use and of possible impacts,
• their information requirements from such a system,
• their own use of such a system (e.g. in their decision-making processes),
• their opinions on the characteristics and design it should have, and
• their risk perceptions on the issue of pharmaceuticals in the environment
The review of stakeholders’ requirements showed two main concepts of what such an information system should contain, and what it should deliver. These have been termed “knowledge base” and “decision support system for doctors / pharmacists / patients”. These concepts do not exclude each other, nor do they have to be independent of each other. Several stakeholders suggested the need for a system that covers both approaches. Key recommendations included:
• An information system on PIE should not be exclusively a decision support system for doctors/pharmacists/practitioners, but also have elements of a knowledge base approach. The stronger the knowledge base component of such an information system, the higher the chances for interesting present and future uses for the information and for positive impacts on the environment.
• Data should not only be complex and aggregated. Simple, “raw” data should also be made available, so as to fulfill the requirements of stakeholder groups beyond the medical system.
• In view of the data and knowledge gaps on the topic and of the importance of precaution in the water cycle, a more ambitious approach to an information system on PIE seems preferable. An ambitious information system on PIE would however require a certain level of resources.
• An independent body in charge of an information system on PIE has strong support among stakeholders, and is seen as key to generate trust in the system. Due to the high complexity of this information and the possibility of errors affecting this trust, data should not just be reviewed for internal consistency, but should be extensively proofed and compared with other available data.
The overall details of the study have been reported in the deliverable D6.1.
A second task was focused on the possible future impact of an European Classification System and the comparison with the Swedish system. First, five qualitative interviews with experts from the pharmaceutical industry, authorities (Stockholm County Council and Swedish Association of Local Authorities and Regions) and science have been conducted in summer 2011. The result was: there is not sufficient quantitative data for an impact assessment of the Swedish classification system.
Other results were that the classification does not seem to be used extensively by individual doctors, mainly because it seems to be not very user friendly. On the other hand, it was thought that the fass.se system is a first and important step to collecting information on the topic and it helps to raise awareness around the issue.
The real impacts are rather expected to occur through the “wise lists”, as county councils increasingly use the information from the classification system to compile those lists, which doctors often consider in their prescription decisions.
A problem seems to be the lack of adequate data. At the time of the interviews, 50% of all substances were not classified. So it is not possible to take Sweden as a quantitative benchmark and to compare it with the results of the following Delphi-Study.
The Delphi method is a structured survey and communication technique, developed as a systematic, interactive forecasting method which relies on a panel of experts. In the Pharmas-Project it was conducted in Germany, the United Kingdom and Hungary. In total, responses were obtained from 186 experts, from various different stakeholder groups which are relevant for the issue of pharmaceuticals in the environment. Besides practitioners, pharmacists, patients and the pharmaceuticals industry, they represent water utilities, medical, pharmaceutical and environmental authorities, NGOs and scientists.
Analysis of the survey responses showed that there is a very high approval of the introduction of an EU-wide environmental classification system for pharmaceuticals. However compared to other possible measures, such as intensified water treatment or proper disposal of unused pharmaceuticals, it is not a top priority. Furthermore, experts assume that the classification system will be used by a significant share of relevant stakeholders (for example between 17 and 59% of doctors depending on the country and framework conditions) and that significant reductions of pharmaceutical residues in the water are possible (between 23 and 50% depending on the country and the group of pharmaceuticals). The results of the first round of the questionnaire have been presented in deliverable 6.2.
Finally, applying the results from the studies described above, and by exploiting the experiences with the existing Swedish system, a prototype for an EU-wide risk and hazard classification system for human pharmaceuticals has been developed. The core of the system is a database that comprises and discerns relevant data on cytostatics and antibiotics. The overall result from Tasks 6.1 and 6.2 on the prerequisite of an EU-wide risk and hazard classification system was the requirement of a knowledge based decision support system (DSS) with varying levels of details regarding the environmental risk of the active pharmaceutical ingredient (API). From a simple hazardous phrase to extensive detailed information, such as sales data, environmental distribution, toxicity, biodegradation, bioaccumulation, removal in sewage treatment plants (STPs), chemical and physical properties of APIs. A common statement from stakeholders was that the more solid and strong the foundation of the knowledge base component becomes - the higher the chances for interesting present and future uses for the information and for positive impacts on the environment. An additional requirement from several stakeholder groups was that such in-depth information should be made available to view and to extract in a simple way. The prototype is described in Deliverable 6.3 and is available at http://eupharmas.ivl.se/
Communication
Project communication and result dissemination were addressed in an individual work package (WP7). This section summarises the actions and results of WP7 over the project’s 39 months.
“Pharmaceuticals in the environment” is an issue receiving growing attention. In particular, information regarding the effect on organisms is more and more required by the actors with influence on the life cycle of PPs, which includes regulators and the general public.
The objective of this work package was to expand and disseminate, as widely as possible, the main information on the many dimensions of this topic. In other words, the aims were to:
• increase the visibility of the topic among the scientific and the policy-making communities by fostering communication and collaboration between scientists and regulators, so as to catalyse and promote further exploration, discussion and collaboration on PPs by all stakeholders and beneficiaries of the work.
• To reach out to the general public (as a group with strong possibilities to influence the presence of pharmaceuticals in the environment) by attractive, simple and easy-to-read language.
• To target media by preparing specific supports corresponding to their requirement and to facilitate their distribution
Communication efforts were addressed to a wide range of users (scientists, the general public and media, and policy makers) and promoted in different ways: website and multimedia support (CDRom, interactive web animations), newsletters and leaflet, a policy brief, and events (corresponding to the 5 sub-activities of the WP).
Task 7.1 concerned the development and maintenance of a dedicated website. The structure was discussed among the partners; the actual website implementation was subcontracted by the project’s scientific coordinator. Ecologic Institute developed texts to address the general public and policy-makers. The web site is accessible at the following address: www.pharmas-eu.net. Figure 13 illustrates the website home page. The main structure of the site has been presented and described in Deliverable 7.5.
Figure 13: Pharmas website home page
Task 7.2 was dedicated to the production of printed material: project newsletters (targeting the scientific community and policy makers) and leaflets (targeting the broad public).
A total of 6 project newsletters were developed over the project duration. Armines held the lead responsibility for this tasked, and was assisted by EHESP and Ecologic Institute. The newsletters were 4 pages in A5 size, and were made available for download on the project website. Starting with Newsletter 3, newsletters were also distributed to an email list of over 700 scientists and policy stakeholders involved in the topic of pharmaceuticals in the environment. All NLs were also made available on the project website for download.
NL1 introduced the project, its aims, and the project consortium. The following NLs presented simplified summaries of the scientific results of different work packages/tasks. The release dates of the different NLs was adjusted from the original plan (planned every 6 months), according to the availability of project deliverables and relevant themes. For instance, NL5 was timed so as to summarise the results of the project-organised International Conference on Pharmaceuticals in the Environment, held in early June 2013 in Nimes, France.
A leaflet targeting the broad public was developed in the final project year. At the 3rd Project Meeting (February 2013, Copenhagen) the decision was taken to develop a leaflet targeting the public that is not focused on the project or on project results, but rather on providing simple but accurate information on the topic of pharmaceuticals in the environment. Ecologic Institute drafted a first version of this leaflet, which was distributed to project partners, who provided valuable comments and suggestions to improve the message and its delivery. Finally, Armines and Ecologic Institute translated the English original into French and German. The leaflet will be permanently available over the CD-Rom, has been made available on the project website, and circulated to the over 700 contacts in the project distribution list.
Project newsletters and leaflets are collected in Deliverable 7.1.
Task 7.3 had as aim informing policy-makers and policy stakeholders of project results, by presenting them at meetings, organising events, and using policy briefs and policy-relevant communication channels.
Under WP7 of Pharmas, Ecologic Institute organised a Science-Policy Event in November 2013, which took place in Brussels. The event presented the results of 5 projects on pharmaceuticals in the environment: 2 FP7 projects (PHARMAS and CYTOTHREAT), an Interreg project (noPills), a German Environmental Protection Agency project (Global Relevance of Pharmaceuticals in the Environment), and a UN project (WHO project on greening procurement of pharmaceuticals). DG Research provided the facilities for the event, which was attended by approximately 30 policy stakeholders, active at EU or at Member State level. The documentation of this event (slides, presentation summaries, stakeholder position statements) is available on the project website.
In March 2014, the main results of the PHARMAS project were presented in a meeting of the Working Group E (Chemicals) of the Water Framework Directive’s Common Implementation Strategy. Representatives of the European Commission and of the 27 Member States were given a summary of the final project results.
In addition to the presentation of project results at purely scientific events (which members of WPs 1 to 5 carried out), presentations on PHARMAS were also held at conferences and events which partly attract policy stakeholders, such as the Berlin Water Week 2013, and the 3rd International Conference on Sustainable Pharmacy 2012. Thomas Backhaus (WP3) presented the project at the 5th EUFEPS World Conference, on Drug Absorption, Transport and Delivery, a conference attended mainly by people from pharmaceutical industry. Ecologic Institute presented and disseminated information on the Pharmas project to visitors (policy-makers, scientists and stakeholders) of the 12th edition of Green Week, the biggest annual conference on European environmental policy, which took place on 22 - 25 May 2012 in Brussels (the focus for 2012 water, and the Green Week was titled "Every Drop Counts").
Partners of other WPs were also encouraged to use policy-relevant channels of communication to disseminate their scientific results. The article of Rik Oldenkamp et al. (2013) (WP4) was turned in and featured in the Science for Environmental Policy Newsletter (Issue 332), which enjoys wide readership in policy circles.
The results of the different PHARMAS work packages have been summarised and simplified, so that they are accessible for policy stakeholders. This policy-relevant summary of project results have been made available in the project’s Policy Brief. Dissemination of the Policy Brief is through project website and distribution over the project’s electronic distribution list.
The Policy Brief is presented in Deliverable 7.2.
Task 7.4 concerned the organisation and execution of an International Conference on Pharmaceuticals in the Environment. This event was initially planned for December 2012. After discussions with the Cytothreat coordinators, it was decided to hold a common conference with results from both projects in June 2013 in Nîmes. The Conference was titled “Pharmaceutical Products in the Environment: Is there a Problem?”, and gathered 110 participants coming from USA, Canada and Europe. Eastern European countries were strongly represented thanks to CYTOTHREAT partners and their networks.
A keynote lecture, two plenary lectures, 24 oral presentations and 35 posters were presented during the two days conference. It was split into 3 sessions:
1. Actions at the European level
2. Environmental and Human Risk Assessment of Pharmaceuticals
3. Perspectives on Problem Assessment
Backdrop to the discussions was the substantial amount of work on the topic carried out during the last 15 years. Over this period, the scientific, regulatory and industry communities have performed research on occurrence, fate, effects and risks of pharmaceuticals in the environment. Regulations have also been developed on the risk assessment of environmental exposure to these compounds. The conference focused on the following questions: What are the lessons learned from all this new knowledge? Can they guide the development of future research and policy initiatives? And is there a real problem with pharmaceuticals in the environment?
Task 7.5 covered the use of multimedia to disseminate the project topic and project results.
Under this task, Ecologic Institute developed a web animation on the topic of pharmaceuticals in the environment. This was the main means to reach out to the general public on this topic. The animation was produced between September 2012 and May 2013, and publicly launched at the above-mentioned International Conference on Pharmaceuticals in the Environment, held in Nimes on June 2-3, 2013.
The idea to use a web animation to communicate the science on the topic to the broad public was due to the possibility of using graphic illustrations to visualise issues that are otherwise hard to visualise, the possibility of attractive visual and sound effects, as well as the possibility of communicating the topic in a didactic and entertaining way. The animation’s storyline has the following structure:
1. Take up some common myths about the impact of pharmaceuticals on human health, poke some fun at them.
2. Dismiss these as not scientifically proven, current scientific consensus is that this subject poses no risk to human health.
3. However, there are instances of significant impacts of drugs on the environment: mention examples.
4. Explain how pharmaceuticals make their way into the environment.
5. Highlight actions being taken by science and the policy world.
6. Highlight actions that consumers can take to reduce their own input of pharmaceuticals into the environment.
The animation’s full title is “The Drugs We Wash Away: Pharmaceuticals, Drinking Water, and the Environment”. It can be viewed on YouTube under the following link: http://www.youtube.com/watch?v=OYbRlJLBzn4 . The INTERREG project “noPills” and its partner project “Den Spurenstoffen auf der Spur” (DSADS), working on the same topic in Germany, produced a German version of the animation. This version is available under: http://www.youtube.com/watch?v=6blafjHQGvw .
At the moment of writing, the English version has 9,506 views, whereas the German one has 1,011 views, thus summarising over 10,000 clicks, a very good result for this type of animation.
Furthermore, noPills and DSADS have developed a follow-up animation, using the same animation studio, visual and sound styles, and characters of “The Drugs We Wash Away”. This animation, also available in English and German, was launched at the end of March 2014, and is available under: http://www.youtube.com/watch?v=Y_-DBrrCDA0 and http://www.youtube.com/watch?v=zi_qfXxyEyI .
A detailed report on the animation and its production process is given in Deliverable 7.3.
A CD-Rom was produced to help disseminate the scientific results and ensure the legacy of the project. The CD-Rom follows the format of a webpage (its Start Page opens in a web browser), and provides a complete overview of project, project objectives, project structure, results of the events organised within the project, project consortium, as well as including all project deliverables. The navigation and structure are intuitive and easy to follow.
The CD-Rom is Deliverable 7.4.
List of Websites:
www.pharmas-eu.net
* The project focussed on two categories of pharmaceuticals: antibiotics and anticancer drugs.
* All available information about a selection of these pharmaceuticals was entered into a database.
* Modelling was used to predict the likely concentrations of both antibiotic and anticancer drugs in rivers throughout the European Union; predicted concentrations were in the ng/litre range.
* Measurement of actual concentrations in both surface and drinking water confirmed that concentrations were low, and sometimes not detectable.
* Food, as well as water, was identified as a potential route of exposure to people.
* Toxic effects in a range of aquatic organisms occurred only at high concentrations of cytotoxic drugs - well above environmental concentrations.
* Theoretical work suggested that because many different antibiotics and anticancer drugs are present simultaneously in the environment, then the risk they pose should be based on realistic mixture scenarios.
* Wastewater treatment technologies were shown to degrade parent pharmaceuticals into a range of transformation products, many of which could not be identified or their potential toxicities assessed.
* Probabilistic risk assessment techniques were employed to identify locations in Europe showing the highest human and ecological risks due to the presence of antibiotics and anticancer drugs.
* A detailed qualitative socio-empirical study of the requirements of different stakeholders identified the information they sought.
* Although lack of appropriate data was identified as a major constraint, nevertheless a prototype for an EU-wide risk and hazard classification system for pharmaceuticals was developed.
* It was concluded that the cytotoxic drugs studied are unlikely to pose a serious threat to the environment. The situation is less clear for antibiotics, due to the concern about induced antibiotic resistance in microbes and the potential transfer of this resistance to pathogens.
* Human health is unlikely to be affected adversely by environmental exposure to the drugs studied.
Project Context and Objectives:
Pharmaceutical products (PPs) in the environment represent an increasing concern for environmental scientists as well as regulatory institutions. Some PPs and their metabolites are not removed during conventional biological treatment and enter the water supply via wastewater-treatment plants (WWTPs) and can also reach drinking waters. Much has been published concerning the occurrence, fate and behaviour of PPs, but risk assessment for the environment and for human health still remains weakly investigated.
According to the international literature, it appears that most of the up to now investigated pharmaceuticals do not pose an acute threat to the environment, but that a small number do, and that none appear likely to pose a significant threat to human health (via environmental exposure). These conclusions are supported by a recent and detailed human health risk assessment (Cunningham et al., 2009 ). However, this apparent certainty that there is no human health risk posed by the presence of human pharmaceuticals in the environment is not supported by deeper thinking. For example, the apparently reassuring risk assessments of Cunningham et al (2009) are based on potential risks to healthy adults: the more sensitive and hence vulnerable foetus, children, the elderly and the infirm were not considered. The real situation is that, currently, there are many more uncertainties than certainties, which leaves scientists, the public and the press still unconvinced that drinking water containing a tiny quantity of a pharmaceutical is completely harmless.
The Pharmas project was launched in this context by adopting an interdisciplinary approach to improve the knowledge on risks for human health and ecosystems, due to exposure to pharmaceuticals and their derivates which are released into the environment. An additional key point of the proposal is to include the main transformation products (metabolites, degradation and treatment by products) with the target molecules. The characterisation of such risk has to be unquestionable, robust and reliable. Furthermore, for decision makers and also for the general public, it is important to make this science easily understandable and available.
As it would obviously be impractical to study all pharmaceuticals, or even a reasonable proportion of them, Pharmas chose to investigate antibiotics and anticancer drugs because they are considered most likely to be a threat to the environment and/or human health. Anticancer drugs can be cytotoxic, genotoxic, mutagenic and teratogenic, and it is generally accepted that no threshold of safety can be given for some of them. Moreover, this group of pharmaceuticals has been studied very little. Concerning antibiotics, although their presence in the environment has already been well studied, the extent to which they are responsible for the induction of genetic resistance is still not well understood.
The final objective is to reduce the level of uncertainty in the current risk assessment procedure. PHARMAS develops a general procedure taking into account scenarios of exposure, as well as the impact of the chosen PPs on organisms, in order to define the range of possible risks for these populations. The project also covers toxicity of mixtures and transformation products. Moreover, social and policy considerations have been added in order to, respectively, inform the public and to support appropriate changes in regulation.
The Pharmas consortium consists of 13 partners from a total of 6 European countries. The project employs a multidisciplinary approach to the problems associated with PPs in the aquatic environment. The partnership involves partners with complementary skills and competencies. This consortium has been designed to meet the research objectives of this collaborative project.
The project was planned over a period of 36 months, the expected results complying with the global objectives of the project were:
- to determine the human and animal exposure to the target molecules by measuring and modelling the concentrations found from the resource (mainly surface water) to the tap (drinking water)
- to provide the missing data on ecotoxicological and toxicological thresholds of concern, by the implementation of targeted ecotoxicological experiments, the compilation of all available literature and an assessment of the sensitivity of particular populations (embryos/newborns)
- to investigate the toxicity of realistic mixtures involving antibiotics and anti-cancer drugs in several dedicated case studies
- to produce probabilistic estimates of the risk caused by the exposure of wildlife and humans by analyzing the uncertainty and variability in the exposure estimates
- to evaluate the usefulness of ecological effect data for the assessment of human effects and to explore options to develop a common effect endpoint for human and ecological risks
- to identify stable transformation products (through treatment processes) and derive very first knowledge how to render future pharmaceuticals better biodegradable in the aquatic environment
- to develop a web-based system primarily designed as an easily accessible decision support tool for practitioners (e.g. physicians and pharmacists) providing straightforward action alternatives
- to study perception of the topic and improve information to the public
- to stimulate discussion by all stakeholders and beneficiaries of the work by supporting communication and collaboration between scientists and regulators
- to contribute to an evolution of the regulation and EU policies by distilling the policy-relevant information, products and conclusions of the project for the relevant policy fields
To help Pharmas in its objectives and works, the project was structured (Figure 1) around 8 workpackages: 6 technical, 1 for dissemination and 1 for coordination. An Advisory Group, composed of a panel of experts, representative of several stakeholders, has been formed to comment and improve the strategy of the project and its results.
Figure 1: Pharmas organisation
Project Co-ordination
The project overall ran extremely smoothly, with excellent co-operation between partners and frequent opportunities to meet in person and discuss any issues of concern. Project meetings were held annually where at least one representative of each partner was present on all occasions. In addition (sometimes in conjunction with the annual meetings), satellite meetings were held sometimes to discuss particular WPs and sometimes with all WP leaders (the ‘scientific council’) present. A separate ‘Advisory Group’ meeting was also convened to discuss stakeholder issues, at which all of the regular Pharmas stakeholders were present as well as several additional interested parties (including Helen Clayton, DG Env) [see deliverable 8.3]. Thus we attempted to maintain full communication links with stakeholders and to encourage them to present their views on the project.
The scientific council was composed of the workpackage leaders and aimed to manage the scientific activities of the project. Six scientific council meetings were organized during the project: February 21-22rd 2011 (at the KoM), November 25th 2011, February 9-10th 2012 (at the annual meeting), May 23rd 2012, February18-19th 2013 (at the second annual meeting) and March 11-12th 2014 (at the final meeting). The kick-off meeting was a good opportunity to launch the project, to establish the mode of functioning of the project hoped for by the coordinator and to define the role of WP leaders; the annual meetings and final meetings allowing to overview the results of the project and to discuss interactions with other WPs. At these meetings, WP leaders also presented their WPs to the advisory group. The other two Scientific Council meetings involved WP leaders only, and allowed discussion on work progress, administrative and financial aspects, and to envisage future works. In no case was a delay in a specific deliverable thought to cause a negative impact on the progress of other areas of the project.
Project Results:
C. A description of the main S&T results/foregrounds
NB/ Figures referred to in this section can be found in the attached pdf file which contains the text and figures of the entire report prepared by the consortium.
Pharmaceuticals in the Environment raise a number of questions about the risks they represent for wildlife and humans. Currently, the risk assessment approaches are not able to give us a strong and definitive answer because many uncertainties still exist. In the Pharmas project, the workpackages were designed to improve this situation by gathering information on exposure, effects (single and mixture), presence and effect of transformation products for the compounds of greatest concern. The following section reports the main results obtained during the project.
The project was focussed on two categories of molecules of high concern: antibiotics and anticancer drugs. Among the wide amount of such substances, a selection of 19 molecules was made, based on the usage, the available literature and the analytical possibilities:
- Anticancer drugs: 5-Fluorouracil (5FU), Capecitabine (CAP), Ifosfamide (IFO), Cyclophosphamide (CPA), Hydroxycarbamide (HYD), Carboplatin (CAR), Methotrexate (MTX), Tamoxifen (TAM) + Imatinib was at least in WP5 investigated as an example of a new drug, see also C1.1.2
- Antibiotics: Chlortetracycline, Doxycycline, Oxytetracycline, Tetracycline (TET), Cefuroxim, Ceftriaxone, Trimethoprim (TMP), Erythromycin (ERY), Ciprofloxacin (CIP), Levofloxacin, Ofloxacin.
All the available data concerning these molecules were collected from the scientific literature, medical/phamaceutical literature, at both national and international levels but also from available data bases (MISTRA, KNAPPE data base, ERAPHAR database, pubchem substance, ChemBank, drug product data base of Health Canada …) and from the manufacturers.. They included : physico-chemical properties (solubility; adsorption; structure), behaviour in humans and animals (metabolism data, dose-response data in various species, preferably a selection of representative aquatic species (algae, Daphnia, fish, etc) and rodents (for extrapolation to humans)), data on mode(s)-of-action, data on potential interactions, contraindications, characteristics of usage by different countries (prescription; consumption; therapeutic uses), environmental fate and behaviour (emission, occurrence; degradation, transfer) and they are described in detail in the deliverable 1.1.
C.1. Exposure assessment:
Calculation of exposure is a determinant element for a reliable risk assessment for both human and environmental ecosystems. Modelling and real measurement have been implemented in the project on several of the molecules of interest.
C.1.1. Modelling
In order to consider several way of exposure, modelling was performed on water but also on foodstuff.
C.1.1.1. Modelling environmental concentrations of exposure via drinking water
In a first step, the work focused on the potential concentrations expected in sewage effluent and surface waters and established the final predicted values for the cytostatic drugs (5FU, CAP, CP, and CAR) and the antibiotics (CIP, SUF, TRI and ERY) for the different countries in the European Union. To examine potential concentrations of these chemicals throughout European surface waters, the geographic-based water resources model GWAVA run in a water quality mode was used. This model uses geographic data on the location and size of the human European population and their association with STPs. The version of GWAVA used here incorporates a newly available and extensive dataset (2009-2010 information) of locations and number of people connected to sewage discharge points in Europe. Information used in the model included flows through STPs, the hydrology, drug consumption, excretion, removal in sewage, and in-stream half- life, ... The model calculates the water concentrations of chemicals throughout European water courses using an overlay of 177,470 grid squares (cells) of approximately 6 x 9 km (5 by 5 Arc minutes) dimensions. Details of each parameter are further explained in deliverable 1.2.
There are uncertainties in the model parameters determining effluent concentrations, which are critical in estimating river concentrations. In order to assess the impact of this uncertainty, a series of scenarios were run to establish the range of likely river concentrations (and hence likely EQS exceedence), based on the reported literature values. These scenarios were a best case - low excretion, high sewage removal and high in stream dissipation; a worst case - high excretion, low removal and slow in stream dissipation and an expected case, which used weighted average values for these parameters. The expected scenario based on the mean literature values gave reasonable agreement with the few available measured data (detail in the deliverable 1.3). The good correlation between values predicted in this study and those observed in sewage effluent are encouraging and further refinement of the model will be necessary in the future as more information on the fate and behaviour of these compounds becomes available.
Figure 2 presents an example of predicted concentrations for cyclophosphamide in European rivers.
Figure 2: Predicted cyclophosphamide (CP) concentrations in surface water based on mean excretion rate, mean sewage treatment removal, and 50%ile flow across the European Continent taking into account differing national per capita consumption and wastewater discharge values from the GWAVA model.
Considering the other cytostatics, there was a surprising difference in the popularity of these drugs across European nations. The predicted mean effluent concentrations ranged from 2 to 40 ng/L for CP, 0.8 to 2.5 ng/L for carboplatin, 0.3-2.5 ng/L for 5FU, and 8.5-87 ng/L for capecitabine. By linking with the geographic based water quality model it is expected that the majority of European rivers would have concentrations below 1 ng/L for these cytostatic drugs that are considerably below concentrations so far reported to have effects on aquatic wildlife. There does not appear to be any widespread threat to European aquatic wildlife based on our current knowledge, because even in the 90%ile prediction around 80% of European surface waters predicted concentrations for these drugs of (generally) below 0.1 ng/L. The issue of water abstraction for drinking water and foetal health may still require further research. Given its potentially high effluent concentrations and good oral absorption by humans, capecitabine certainly seems worthy of further environmental research.
The GWAVA model was also used for antibiotic modelling in surface water. The two scenarios modelled were expected and best case. The first task was to provide a concentration map for each compound using only the best case scenario (annual average). Figure 3 illustrates the case of Trimethoprim.
Figure 3: Predicted trimethoprin (TRI) concentrations in surface water based on best case scenario (lowest excretion rate and highest sewage treatment removal) and mean flow across the European Continent taking into account differing national per capita consumption and wastewater discharge values from the GWAVA model (little used in Spain).
With the expected scenario at the highest end of results (90%ile) antibiotic concentrations could reach 10s to 100s of ng/L in 10% of river lengths (see deliverable). If we assumed the worst case situation of high excretion, poor removal and low flow then concentrations could reach 100 ng/L in up to 10% of river lengths for all the antibiotics except ciprofloxacin. The most toxic of the antibiotics appears to be ciprofloxacin but under most circumstances all European rivers are around 2 orders of magnitude below the effect level. Trimethoprin appears to be of the least concern having around 6 orders of magnitude safety factor. We cannot comment on to what extent such river concentrations might stimulate antibiotic resistance.
The results of GWAVA models for several European countries is reported in the deliverable 1.4
C.1.1.2. Identification of possible routes of exposure via foodstuffs
Drinking water is not the only route of exposure of human to pharmaceutical products. Food was also investigated as potential vector of contamination.
Indirect human exposure was assessed via the exposure routes from wastewater to rivers and to agricultural soils, and further via bioaccumulation in fish and crops to humans (Figure 4).
Figure 4: Emission pathways and environmental compartments considered in the study.
The model Activity Simple Treat was used. This is a version of Simple Treat adapted to deal with ionisable chemicals. The stated variable is not concentration nor fugacity, but the activity of the chemical. A dynamic plant uptake model for ionisable compounds was developed, which combines the standard model approaches with the concept of the cell model. To validate this model, a literature search for measured concentrations in plants has been undertaken.
The chemical data set (antibiotics and cytostatic drugs) included polar neutral substances, mono- and bivalent acids, mono- and bivalent bases, zwitterions of various types, and metallorganics. This made chemodynamic calculations challenging. A major effort was to modify approaches, among them SimpleTreat, regressions for bioconcentration in fish, and dynamic models for transport of water and substances in soil and plant, plant uptake models, so that multispecies calculations for the various ionic species could be undertaken. For the bioaccumulation in fish and plants, varieties of the cell model for each type of electrolytes were established because existing regressions for the bioconcentration (BCF) are not applicable for many of the investigated compounds. Usual models focus on partitioning between lipids and water. That process is of little relevance for the very polar medical drugs, such as zwitterionic antibiotics (for example, ciprofloxacin, tetracyclines). As novelty we added the adsorption to proteins into the models. The model system was consistently parameterized for the Danish scenario (but can be adapted to any other European scenario).
The results (Table 1) indicate that amendment of agricultural fields with sewage sludge and the subsequent translocation into food crops is most relevant for indirect human environmental exposure to pharmaceuticals, more than ingestion of (river) fish. Highest dietary exposure with root crops (incl. potato), leafy vegetables and fruits (incl. cereals) was predicted for the antibiotics tetracycline, ciprofloxacin, trimethoprim, and others and for the anticancer drug capecitabine and imatinib (all > 1 g/d). Plants generally concentrate polar and nonvolatile compounds (if they are sufficiently persistent to reach the plant), while fish accumulate more lipophilic substances. Most of the investigated compounds were ionizable and thus very polar, only tamoxifen belongs to the very lipophilic compounds.
Table 1. Dietary uptake of antibiotics by women in Denmark, 50%-percentile. Unit ng/d.
women ng/d root crops leafy vegetables fruits & cereals sum crops fish
Ceftriaxone 0.24 0.04 0.01 0.30 0.16
Cefuroxime 0.00 0.00 0.00 0.00 2.3
Chlortetracycline 0.00 0.00 0.00 0.00 0.00
Ciprofloxacin 175.6 38.8 21.4 235.7 7.5
Doxycycline 301.9 14.8 7.3 324.0 53.6
Erythromycin 37.9 156.4 29.5 223.8 0.6
Levofloxacin 34.9 16.9 7.2 59.0 5.9
Ofloxacin 7.5 0.3 0.4 8.3 2.2
Oxytetracycline 0.65 0.16 0.11 0.91 0.22
Tetracycline 650.4 5.0 1.2 656.5 36.2
Trimethoprim 158.3 841.7 164.3 1164.3 5.15
A comparison of predicted to measured concentrations in soil and plant was satisfactory in most cases. However, the variation of measured data is quite high. The BCFs (bioconcentration factors) leaves to soil of trimethoprim, erythromycin and ceftriaxon are above 1 kg:kg. Accumulation in corn is one to two orders of amount lower. Practically no comparison to experimental data could be done for the bioconcentration in fish. Therefore, the results possess therefore high uncertainty and must be considered as preliminary.
In summary, it appears that dietary exposure to human antibiotics via indirect environmental exposure is higher for the pathway sludge - soil - crops than for the pathway effluent - river – fish. In addition by using Danish data, the dietary exposure to human antibiotics via indirect environmental exposure was shown to be particularly high (close to 1 g/d) for several antibiotics, namely the compounds trimethoprim, tetracycline, ciprofloxacin, doxycycline, erythromycin, and a few anticancer drugs (imatinib, capecitabine). However, there remain considerable uncertainties. The deliverable 1.5 details the whole works of this section.
C.1.2. Measured Environmental Concentration
Part of the project was dedicated to real measurement of the chemicals of concern in various water matrices, in order to evaluate the degree of contamination of antibiotics and anticancer drugs, in comparison to other pharmaceutical products.
In a first step an interlaboratory exercise on ATB was organised. 14 laboratories participated including 2 members of the Pharmas consortium. Considering the different profiles of the participants (obtained by a questionnaire) and the capacity of the host laboratory to produce the reference materials, the following antibiotics were chosen: Ofloxacin, Ciprofloxacin, Sulfamethoxazole, Trimethoprim, Erythromycin. 2 matrices were investigated: surface and drinking water. Tests of stability, preservation and transport conditions have been performed. No standard protocol was imposed, neither for the extraction/concentration step nor for the analytical methods/devices.
A total of 77 samples were analyzed to determine concentrations of selected antibiotics and 611 results (including parallels, excluding
In a second step, analytical campaigns were performed. More precisely, several studies were performed in France and The Netherlands that aimed to evaluate the level of contamination of both raw (resources used to produce drinking water) and drinking waters and consequently the level of exposure of the population. Several analytical measurements were performed usingdrinking waters and their resources, by the laboratories of the consortium. Two series of studies were done in France and in The Netherlands.
A selection of target molecules was analysed, which include antibiotics and anticancer drugs selected in the project but also additional antibiotics and other pharmaceuticals (included in the multicomponent analytical methodologies employed). Table 2 presents the molecules analysed for this task.
Table 2: List of pharmaceuticals analysed
Molecules of interest for the project:
Ciprofloxacin Chlortetracyclin Erythromycin Ofloxacin
Oxytetracyclin Sulfamethoxazole Tetracyclin Trimethoprim
Cyclophosphamide Imatinib 5FU
Additional antibiotics:
Danofloxacin Difloxacin Enoxacin Enrofloxacin
Flumequin Lomefloxacin Nalidixic Acid Norfloxacin
Oxolinic Acid Pipemidic Acid Sulfamethazin Ofloxacin+Levofloxacin
Sarafloxacin
Additional molecules:
Atenolol Cafeine Carbamazepine Codeine
Iopromide Morphine Oxazepam
Sorafenib Erlotinib Sunitinib
Results from France showed only 1 positive result – antibiotic at 17 ng/L - out of more than 500 analyses. Therefore, considering the exposure dose calculation and the concentration found in drinking water, it can be concluded that human exposure to antibiotics and anticancer drugs via drinking water is unlikely. Considering other pharmaceutical products (PPs) analyzed in this work, the maximum concentration in drinking water was found to be 37 ng/L for Iopromide. Even present in drinking water (DW), such concentrations corresponds to an exposure to 1 ng/ kg bw/ d, 3.7 ng/ kg bw/ d, 5.5 ng/ kg bw/ d respectively for adults, children and infants.
Concerning experiments undertaken in the Netherlands, a smaller survey was carried out including surface water, groundwater and drinking water samples on a limited number of targets: 3 classical antibiotics and 5 anticancer drugs including 4 protein kinase inhibitors (PKIs) (Imatinib, Sorafenib, Sunitib and Erlotinib). The concentrations measured were found to be higher in sewage effluent than in ground and drinking water. Maximum concentration for antibiotics in surface water was quantified at 19 ng/L. PKIs were measured in surface water and even in drinking water but in the last case at concentration below 1 ng/L. Deliverable 1.7 reports in detail the results of these analytical tasks.
C.2. Effect assessment
For many of pharmaceuticals detected in the aquatic environment, the possible impacts on the environment and public health are unclear. Indeed, very few data on ecotoxicology are currently in the public arena, either for single substances (in particular for cytotoxic drugs), or mixtures.
This part of the project aimed to provide some answers as to whether these drugs might be of concern at the concentrations at which they are detected in the environment.
It seems to be widely accepted that environmental exposure of humans to pharmaceuticals (in drinking water, for example) will have no effects. This is because the exposure dose (in food or drinking water) is likely to be far lower than the therapeutic dose (the one given to patients). But in reality there is very little information to enable us to either confirm or refute this hypothesis. One concern is that there may be particularly sensitive sub-groups of humans (e.g. foetuses, or the elderly and infirm people).
In (probable) contrast to humans, it is known that environmental exposure to pharmaceuticals can adversely affect wildlife. Nonetheless, at the time of writing the Pharmas proposal, essentially little is known about possible effects of antibiotics and anti-cancer drugs on aquatic organisms. We aimed to fill that knowledge gap with a focus on the effects of anti-cancer drugs on fish, invertebrates, and algae.
Finally, concern has been expressed that the human foetus and/or young child may be more sensitive than the adults of the species. Although we clearly cannot test this hypothesis using human subjects, the project aimed to compare the sensitivities of young and adult fish. Fish are often good surrogates for mammals (see, for example, Zelikoff, 1998 ; Bols et al, 2005 ), so we anticipate that the results can be extrapolated to human health (for risk assessment purposes).
In this context, works on single substances were to (i) determine the minimum doses of anticancer drugs that cause measurable effects on humans with particular emphasis on sensitive sub-groups such as the foetus, (ii) assess the genotoxic effects of some anticancer drugs on fish, an aquatic invertebrate and algae, and (iii) compare the sensitivities of embryo and adult fish to some anticancer drugs.
The assessment of the effects of cytotoxic drugs in humans was not easy to perform due to the limited data available at public area. However, the topic was addressed and reported in deliverable 2.1. In that report, the main conclusions were:
- For different pharmaceuticals the difference between the daily therapeutic dose and the likely level of intake via drinking water and/or fish consumption ranges from two to ten or more orders of magnitude.
- All assessments published to date have concluded that the risk to human health from the presence of pharmaceuticals in the environment is very low, if not negligible.
- There are some uncertainties associated with this conclusion of very little or no risk, but these are unlikely to alter that conclusion.
- Particularly sensitive sub-groups of people may exist, but even these are unlikely to be adversely affected by the presence of human pharmaceuticals in the environment.
- Cytotoxic drugs are a special case because they are designed to be toxic. Although environmental concentrations of these drugs are currently largely unknown, they are very unlikely to be high enough to cause adverse effects to people inadvertently exposed to them.
- Reassurance for the conclusion of little or no risk to human health is provided by the fact that the health of the foetus is not affected if its mother undergoes chemotherapy for cancer during the second or third trimester of pregnancy (see a recent report on this subject by Amant et al, 2012).
The potential for effects of environmental exposure to cytotoxins on foetal health during the first trimester of pregnancy are less clear, but given that :
a) there are expected to be only very low concentrations (at the absolute most in the ng/L range) of cytotoxic drugs present even in surface water;
b) that concentrations in treated drinking water are likely to be lower still (and were undetectable in most of the samples analysed for this project – see above conclusions on measured and predicted concentrations of these compounds in surface and drinking water);
c) that even the most potent cytotoxin tested here (cisplatin) only induced toxic effects at several hundred µg/L;
- we think that the exposure levels via drinking water even to this vulnerable group will not be at concentrations high enough to induce adverse effects.
With particular reference to the potential for increased sensitivity of human sub-groups (e.g. exposure of the foetus during pregnancy, or the elderly/infirm), it was concluded that although an extremely ill person might react badly to a pharmaceutical product prescribed to them, it seems unlikely that amounts present in drinking water would be sufficient to trigger adverse effects. Further, a recent study (Amant et. al., 2012 ) has shown no effects on foetuses of pregnant mothers administered chemotherapy during the second or third trimester of pregnancy, suggesting that the much lower doses of cytotoxic drugs present in drinking water are very unlikely to affect the unborn foetus.
The assessment of genotoxic effects was investigated. Although a number of cytotoxins were assessed in some assays, the main focus was on three drugs, namely 5-Flourouracil (5-FU), cyclophosphamide (CPA) and cisplatin (CIS). Two immunocytochemistry assays (using antibodies to H2AX and RAD51, which are DNA repair enzymes) were originally selected for development in fish and algae, but they were found to be unreliable and unrepeatable using the tissues collected. A considerable amount of time was spent trying to develop this method, because it was felt that if a reliable assay had been forthcoming it would have provided invaluable and unique information on the toxic effects of these drugs in fish. As it was unsuccessful, it was decided to focus on the toxic effects of cytotoxins in algae, daphnia and adult zebrafish. A considerable amount of data describing the toxicity of these drugs in zebrafish embryos was produced. In addition, the measurement of the expression levels of genes encoding DNA repair enzymes (specifically, RAD51), and also the tumour suppressor gene P53, (which can be activated in response to DNA damage) in exposed zebrafish was performed in both adults and embryos.
In general, ecotoxic effects of cytotoxic drugs in all organisms studied (algae, daphnia, zebrafish adults and zebrafish embryos) were found to occur only at extremely high concentrations (several hundred micrograms or even several milligrams per litre), far higher than would be observed in European aquatic environments. This does not eliminate the potential for ecotoxic effects in areas where mixtures of these chemicals are present, but it does suggest that individually these drugs are less likely to cause harm to aquatic organisms, than we might have expected given their known toxic nature, particularly at the concentrations predicted to be found in surface waters, (albeit that toxic doses used in humans are at higher concentrations than those used in this project).
Cisplatin was found to be (relatively speaking) the most potent toxin in both the daphnia and zebrafish embryo assays. This compound was also tested in a pilot study at a single concentration alongside 5-FU and cyclophosphamide using adult zebrafish, however the cost of cisplatin excluded it from a comprehensive dose-response study with adult zebrafish (Partner 1 uses a flow-through exposure system, so several thousand euros worth of the chemical would have been required to conduct a study using this system). Nevertheless, this compound was only toxic in the assays performed here at concentrations exceeding those reported in the environment; a concentration of several hundred µg/L cisplatin was required to elicit a response in the zebrafish embryo toxicity assay, and in the daphnia immobilisation assay the EC50 of this chemical was 2.06 mg/L.
Algal inhibition assay
Inhibition of growth of algae (Pseudokirchneriella subcapitata) was assessed using kits purchased from MicroBioTests, which enable the calculation of a 72 hr EC50. Briefly, test organisms (Pseudokircheniella subcapitata) were stored immobilised in ‘algal beads’ and were de-immobilised on the first day of the assay. Algae were incubated in medium containing the toxicant at 22±1°C, under constant uniform illumination, for 72 hours. OD was determined using spectrophotometer after 24, 48 and 72 hours incubation. The three cytotoxins selected for use in this assay were found to have relatively high EC50 values in this assay (i.e. they are not considered to be highly toxic, and do not show effects in this assay at levels observed in the aquatic environment); these values concur with those reported by Zounkova (2007) in a similar algal growth inhibition assay (both in terms of rank order of potency and, broadly speaking, absolute potency). The EC50 values of these compounds were as follows: 5FU: 0.44 mg/L; CIS: 7.4 mg/L; CPA:3179 mg/L (see Figure 5).
Figure 5: Results of the algal inhibition assay undertaken with cytotoxic drugs
Daphnia immobilisation assay
Daphnia immobilisation was assessed using kits also purchased from MicroBioTests. Briefly, ephippia of Daphnia magna were hatched under optimal conditions (6000 lux; 22°C), a process which takes approximately 72 hours. Once hatched, and with 24 hours of hatching, neonates were fed with an algae supplement, then transferred to test plates where they were separated into replicate wells containing the exposure substance at a range of concentrations. This kit allows for 5 different treatments plus a control, with 4 wells per treatment and 5 neonates in each well. Daphnia were incubated in the exposure medium and were maintained in the dark at 22°C; numbers of dead and immobilised organisms were assessed after 24 and 48 hours. In all cases, a range finding test was performed in the first instance, followed by at least two definitive tests. Results obtained for the three cytotoxins tested here showed the same rank order of potency as those reported in Zounkova (2007). That is, in this assay, cisplatin was the most potent compound. This is different than the order of potency observed in the algal tests where 5-FU is consistently observed to be the most potent of the three compounds. In both daphnia and algae, cyclophosphamide was found to be the least potent of the three compounds and in fact concentrations of several g/L were required to reach induce a response of 50% (Figure 6). The EC50 values (based on a mean of two repeat assays) of the three cytotoxins tested were as follows: CIS: 2.06 mg/L; 5FU: 243 mg/L; CYP: 2050 mg/L.
Figure 6: Effects of three cytotoxic drugs on immobilisation of Daphnia magna
Zebrafish assays
The zebrafish (Danio rerio) is an established model system in cancer research. It is used to study cancer development and progression and applied in screens to detect new anticancer drugs (Feitsma & Cuppen, 2008 ). Additionally the zebrafish is also commonly used in (eco)toxicology studies. Established OECD guidelines are available for an adult prolonged toxicity test (with a 2-week duration) (OECD 204); more recently, OECD test-guidelines were developed using zebrafish embryos for toxicity testing (OECD 236). As part of the Pharmas project we were investigating the effects of anti-cancer drugs on vertebrates. Cancer is defined as the uncontrolled fast growth of abnormal cells in the body. Most anti-cancer drugs disturb mechanisms of cell proliferation in order to stop the cancer cells from growing. During early development, normal cells are also proliferating very fast and disturbances during this period can have detrimental effects on development and pregnancy outcome. Therefore exposure to anti-cancer drugs during early development might be of higher risk. We investigated these potential risks by studying the effects of widely used anti-cancer drugs on the early embryonic development and on the larval stage of zebrafish and compared them with effects observed in adult zebrafish.
Adult zebrafish exposures
Adult zebrafish exposure studies were undertaken using a flow-through exposure system. Tanks of mixed-sex fish (6 males plus 6 females) were exposed to the test substance for 14 days according to OECD guidelines 204. A range of physiological end points were measured and the results assessed for all individuals and within the two sexes.
An initial pilot study was undertaken using three cytotoxic drugs, based on highest observed environmental concentrations, as well as the results obtained in the embryo toxicity assay undertaken by Partner 9. This pilot study used 10 µg/L cisplatin, 100 µg/L 5-FU and 100 µg/L cyclophosphamide. Following consideration of the results obtained in the pilot study, as well as the fact that cisplatin is an extremely expensive drug to use at high concentrations in a flow-through system, further studies looked in more detail at varying concentrations of 5-FU and cyclophosphamide.
Hepatosomatic index
In the pilot study, the hepatosomatic index (HSI) was observed to be significantly different to the controls when assessed using the student t-test (Figure 7). The livers of exposed fish were significantly smaller than those of the controls for all three drugs, with the most significant difference observed in those fish exposed to 5FU. In subsequent studies, the effect on HSI of exposed fish was not so clear; in one experiment using 5-FU (but where the fish were not as old as in the pilot study) there was no effect and in a subsequent study there was a dose-related trend in reduction of HSI, but the responses in individual treatments were not statistically significantly different from the control, as they had been in the pilot study.
Figure 7: Mean HSI of adult D.rerio exposed to either 100 µg/l 5FU, 100 µg/l CYP or 10 µg/l CIS for 14 days. A single asterix denotes significance at 5% in the students t-test, a double asterix significance at 1% when compared to the controls. Error bars show CI @ 95% Confidence Level.
qPCR
As the anti-cancer drugs are cytotoxic, it was suggested that this reduction in liver size could be the result of a loss of liver cells. In order to assess this, relative expression of RAD51 and p53, genes involved in the repair of DNA damage, in the livers of these animals were measured using quantitative rt-PCR. Primers were obtained from our partners in IVM. Two housekeeping genes were assessed, β-actin and elongation factor 1 α (ef1α). The results were normalised against the housekeeping genes using the Q-gene application, configured for “Mean Normalised Expression”, and tested for significance against the controls using the students t-test.
The data obtained using qPCR in these assays is not clear-cut. In the pilot study there appeared to be a consistent trend of the response of the genes analysed (that is, an increase in expression of RAD51 and a decrease in expression of p53) to all three chemicals tested. This pattern of response was not repeated in the concentration series experiments with 5-FU however (for full details please refer to D2.2). Currently the reasons for this are not clear, but it does seem apparent that there are few significant responses either at the morphological or genetic level in adult zebrafish which have been exposed to cytotoxic drugs, at relatively high concentrations, for a 2 week period.
A full description of the results obtained from these genotoxicity studies is given in Deliverable 2.2.
Zebrafish Embryo Toxicity
In general all tested anticancer drugs showed low toxicity in zebrafish embryos (LOEC at mM (g/l) range). The three compounds tested in more detail induced very specific effects. 5-FU: hyperpigmentation, p53 induction; CPA: short tail phenotype; Platins: none hatching. For CPA and 5-FU the results were similar to published data on rodent models and human studies. Cisplatin was the most toxic drug tested (LOEC 1 μM ). We could not detect effects on genes involved in DNA damage for any of the compounds at the concentrations tested. CPA and Cisplatin are known to bind to DNA and crosslink it. Despite the similar mechanism they induce very different effects in zebrafish embryos and larva. Therefore we conclude that endpoints of DNA damage are not the most sensitive ones for anticancer drug exposures. CPA, 5FU and Cisplatin showed differences in the developmental windows when the embryo is the most sensitive to the drugs. Cisplatin and CPA are more toxic during the early phase of the development whereas 5-FU shows effects at the later stages.
Furthermore, we could not find any literature comparing effects of anticancer drugs at different developmental stages of zebrafish.
For a more comprehensive description of the results obtained from the work undertaken for this workpackage see deliverable report D2.2.
Cisplatin and Carboplatin
These compounds both affected hatching of zebrafish embryos. It is known that platinum substrates can bind to metalloproteases and inhibit their function (Arnesano et al., 2009) . The hatching enzymes of zebrafish are metalloproteases. We could not detect any effects on gene level or visual malformations that might indicate that the excretion of the hatching enzymes is affected. Nevertheless we could clearly see that the chorion of exposed embryos is not digested and therefore prevents hatching. Other studies also detected a postponed hatching when embryos are exposed to platinum complexes like PtCl2 (Osterauer et al, 2011 ). We assume that Cisplatin and Carboplatin and other platinum complexes bind to the hatching enzymes and inhibit their function. The hatching process is highly conserved between different fish species (Sano et al., 2008 ). Therefore the effect of platinum complexes could be a general effect for most fish species.
Cyclophosphamide(CPA)
CPA induced very specific effects in exposed zebrafish embryos. The most sensitive time for exposure was after organogenesis has started. This fits with reports on other species. Also the effects on body size were comparable. The decreased heartbeat could be detected at concentrations where the total length was already reduced. Therefore the size reduction is not caused by a reduced heartbeat. No genes linked to DNA damage were induced. We assume that CPA induces oxidative stress followed by reactive oxygen production and increased oxygen consumption which leads to the observed effects (body size, eye size, heart beat) and cell death (gene induction).
5-Fluorouracil
5-FU was tested at concentrations up to 20mM (near the solubility threshold of 5-FU in water). No teratogenic effects could be detected besides slight hyperpigmentation at levels higher than 10mM. No genes linked to DNA damage were induced. Only p53 was elevated in exposures after 3dpf at 10mM concentrations. P53 induction can lead to hyperpigmentation and might explain what we could observe in exposed embryos. Our findings show that 5-Fu is up taken and can be converted into its active form in zebrafish embryos older the 3dpf. Why we could not detect effects at lower levels and before 3dpf (TS protein increase, 5-FU degradation, low uptake) is still unclear.
Comparison of sensitivities of adult vs embryo zebrafish stages
From the studies carried out with embryonic, larval and adult stages of zebrafish in Pharmas, compound- and life-stage specific sensitivity to the anti-cancer drugs was observed, albeit at concentrations higher than those found in the environment. Cisplatin showed a life-stage specific effect on hatching during early development at the lowest effect concentration of the three compounds tested in embryos (0.001 mM or about 300 µg/l). In adults, 10 µg/l cisplatin did not show any effects on the endpoints measured. Higher concentrations were not tested, making it difficult to conclude if adult or early life stages are more sensitive. For 5-FU, effects on HSI were found in adult fish at lower concentrations than those observed in embryos and larvae, suggesting higher susceptibility of adults to this compound, likely due to the enhanced bioactivation of 5-FU by the adult liver or lack of uptake in embryos. For CPA, a life-stage specific effect on growth during early larval development was observed, though at concentrations higher than those tested in adults. CPA at 100 µg/l induced a significant reduction in HSI, indicating an enhanced sensitivity of adult stages to this drug.
The effects of anti-cancer drugs observed in Pharmas were at concentrations far exceeding environmental concentrations. It should be mentioned, however, that the exposure duration used in this study was rather short, especially those used for embryotoxicity testing, and we cannot rule out that longer term exposures may result in effects at lower concentrations. Indeed, complete life cycle studies with 5-FU in zebrafish revealed genotoxic effects at much lower concentrations than used here (Cytothreat presentation, Brussels 2013). We also focussed on phenotypical effects of anti-cancer drugs on zebrafish development, and have not assessed the long term effects of early-life stage exposure on reproduction, growth or survival.
C.3. Mixture substances
The toxicology and ecotoxicology of chemical mixtures has one strong, common pattern – independent of the exposed organism, analyzed biological endpoint and chemical composition of the mixture: the joint toxicity of a combination of chemicals is in the overwhelming majority of cases higher than the toxic effects of each single compound. This implies that chemical risk assessments and guidelines need to take combination effects into account – or otherwise run the risk of a serious underestimation of the actual risk for humans and the environment.
The classical toxicological concepts of Concentration According (CA) and Independent Action (IA) have been successfully used in a range of scientific studies to predict the (eco)toxicity of various chemical mixtures, including combinations of pharmaceuticals. CA is thought to be applicable to mixtures of similarly acting compounds, while IA is based on the idea that the components in a mixture act independently, via different physiological pathways on a common biological endpoint. CA and IA make use of information on single substance toxicities for predicting the joint action of a mixture.
Consequently, both concepts assume that the environmental or human health risk of a mixture is dependent on the number of present compounds, their individual toxicities and their concentrations in the mixture. Although several antibiotics and anti-cancer drugs have been detected in aquatic systems, neither have the available analytical fingerprints been analyzed from a mixture toxicity perspective, nor has the expected exposure to pharmaceutical mixtures been modeled on a broader scale and then analyzed for its potential risk.
In this context, works have been performed in order to (i) assess the environmental risks of realistic mixtures of antibiotics and/or anti-cancer drugs; (ii) analyze by how much current regulatory single-substance oriented risk assessment strategies run the risk of underestimating the actual risk of environmentally realistic exposure situations; and (iii) provide the necessary scientific tools and regulatory options for considering the joint effects of antibiotics and anti-cancer drugs during the environmental risk assessment process. To reach these objectives, the assessment of realistic mixtures of antibiotics and anti-cancer was carried out on fish (embryos) and selected species of bacteria and algae, on the biodiversity and ecological succession in natural microbial communities, on bacterial communities in STP plants and by the optimisation and validation of predictive mixture hazard modelling approaches
The prediction of risk quotient for environmentally realistic mixtures of pharmaceuticals was performed in several steps. A review of the scientific literature on monitoring of mixtures of pharmaceuticals was conducted firstly to lead to a selection of one particular publication with the broadest range of investigated pharmaceuticals, in particular it is one of the very few published studies where antibiotics were included at least to a certain extent (7 of the 26 studied pharmaceuticals were antibiotics). The ecotoxicological data for all the 26 monitored pharmaceuticals were collected from the peer reviewed literature and publically accessible databases. Data on toxicity and ecotoxicity for mammals, birds, higher plants, fungi and bacteria were only available for a very few compounds and organism groups. Those available were grouped in three groups: algae, invertebrates and fish. Finally, the measured environmental concentrations (MECs) and the compiled ecotoxicological data were used for calculating the toxic units (TU = MEC/EC50) of the individual pharmaceuticals for the three organism groups, and the resulting risk quotients (as sums of toxic units, STU) of the mixtures:
All analyses had to be confined to the summation of TUs (i.e. the mixture toxicity assessment according to CA), as the available data did not allow the calculation of the corresponding IA predictions. We therefore also estimated the maximum possible error that might result from this simplification.
A total of 890 individual ecotoxicity values were collected. The STU for algae and invertebrates is consistently higher than 10-3. This implies that the exposure situation might give reason for concern, if the standard single substance assessment factor of 1 000 for the base set of data (REACH guideline) is applied. In fact, mixture risk quotients are then between 16 (Swedish STP effluent) and 45 (French STP). Whether there is an actual risk for the aquatic environment as a consequence of STP discharge then ultimately depends on the actual dilution of the STP effluent in the receiving river.
It might be argued that the application of CA is potentially overconservative. The strict mathematical relation between CA and the competing concept of IA allows to calculate the maximum possible overestimation by CA (in relation to IA); the factor between both predictions being usually between 1.5 and 3, only in two cases (STU for fish for S1-F and L1-F) the ratio is between 4 and 5.
In view of all other uncertainties inherent in the exposure as well as the effect data, a mis-prediction by a factor of 1.5 to 3 might be regarded as of minor concern. In fact, the ratio IA to CA predicted EC50 is well below 2 for all STPs if algae (the most sensitive organism group) are considered. That is, even if a mixture assessment would be based on IA, the joint toxicity of the pharmaceuticals might still be of concern.
The sum of PEC/PNEC ratios (or MEC/PNEC ratios) can be considered a pragmatic approximation of CA, while a STU based risk assessment is conceptually closer to the concept of CA. A comparison between the risk quotient that results from summing up MEC/PNEC ratios and the risk quotient that results from STUs showed that the resulting ratio is less than 1.3 in all cases, indicating that summing up MEC/PNEC ratios (respectively PEC/PNEC ratios) can indeed be used in a first tier risk assessment of pharmaceutical mixtures. The toxic unit distributions (Figure 8) are specific for each organism group and effluent. However, all distributions have a common shape, which is characterized by the fact that a comparatively few compounds always contribute the bulk of the toxicity. In fact, the first 10 compounds always provide more than 99% of the total STU. The detail of each distribution for all STP tested are described in deliverable 3.1.
Figure 8: Example of Toxic Unit Distribution (relative) for algae and invertebrates for a STP‐effluents
Deliverable .3.3 focused then on providing tools to allow also new users to implement Concentration Addition for predicting and assessment mixture effects. These Mixture Tools provide three sets of tools and analysis instruments for prediction and assessment of mixture toxicities. The first toolset provides a framework for the calculation and visualization of mixture toxicities, based on the assumption that sufficient input information is available for the calculation of the toxic units of the individual compounds. This toolset also provides an overview of the quantitative consequences of synergistic interactions. The second toolset comprises a set of functions that can be used to calculate the predicted mixture toxicity according to Concentration Addition and Independent Action in arbitrary scenarios. The third toolset provides a set of concentration-response functions for estimating ECx values (e.g. EC50's, EC10's), which then can be used for the mixture toxicity assessment. Examples for all toolsets are described in deliverable 3.3 and are accompanying the Mixture Tools collection.
Empirical analyses using a range of mainly bacterial assays were conducted in the context of deliverable 3.2. A mixture of 14 pharmaceuticals detected in the effluent of the Ryverket STP in Gothenburg, which includes 7 antibiotics, was tested for its toxicity towards biofilm communities (periphyton), sampled from a pristine site. The reconstituted mixture did not provoke significant effects on bacterial communities at the effluent concentration. The total mixture causes highly significant effects only at a concentration of 100x times the effluent, while first effects of the antibiotic sub-mixture become visible at roughly 10 times the effluent concentration (EC10=10.3x the effluent concentration). Applying an assessment factor of 10 - which is deemed appropriate, given the fact that the available background data from single species assays indicate that bacteria are indeed the most sensitive trophic level and higher-tier tests with fish or invertebrates might hence not provide substantial new risk-relevant information - results in a final empirical risk quotient for the pharmaceutical part of the effluent of almost exactly one. This fits well with the risk estimate of 20 that was put forward in deliverable 3.1 based on predictive modeling and used an assessment factor of 1 000. The results hence (i) strongly support the current regulatory strategy to base the environmental assessment of antibiotics - and consequently antibiotic-dominated mixtures - on tests with bluegreen algae (EMA, 2006), (ii) emphasizes that local conditions, in particular the dilution factor of the receiving waters, need to be taken into consideration for a realistic risk estimate.
For the work with STP-derived bacterial communities derived from STP effluents the mixture of pharmaceuticals were tested in different concentrations. Different community-level toxicity endpoints like cell viability and multiplication, respiratory activity and metabolic diversity were addressed. Also single species toxicity testing on Vibrio fischeri and Pseudomonas putida were performed. Contrary to V. fischeri, STP derived bacterial communities were not affected by pharmaceutical mixtures in the range of measured effluent concentrations. This could be an indication that these communities are already adapted to mixtures of pharmaceuticals in actual measured concentrations. Details on the experimental work are given in deliverable 3.2.
Tolerance development was further investigated in the work on deliverable 3.5. As a worst case scenario, we investigate the impact of treated effluent from a plant receiving waste water from bulk drug production in Patancheru, India and its main antibiotic ingredient, ciprofloxacin, to natural bacterial communities (periphyton biofilms). Toxic effects were assessed as changes in carbon utilization pattern using Biolog Ecoplates™. Bacteria exposed in Ecoplates directly after sampling from an uncontaminated site were highly sensitive towards both ciprofloxacin and effluent. Observed EC50 values for the average ability to degrade carbon source were 70.3 nmol/L for pure ciprofloxacin , respectively 0.08% for the effluent. The degradation of specific individual carbon sources was even affected already at concentrations a tenfold lower. These concentrations are lower than the effect concentrations observed in previous studies, indicating that natural periphyton might provide suitable test material for investigating environmental effects of antibiotics. The high toxicity of the effluent also clearly indicates the need for appropriate risk assessment and management strategies for the investigated effluent.
Pre-exposure to 0.1% of the effluent clearly induced tolerances both to the effluent itself, but even more so to pure ciprofloxacin. The EC50 of ciprofloxacin was increased by a factor of al-most 4. This substantial development of tolerance against ciprofloxacin caused by the effluent indicates the possibility of resistance development in the environment that might potentially hamper the use of fluoroquinolones in human medicine.
C.4. Transformation products
After usage, pharmaceutical products will undergo transformations starting from human metabolism to degradation in conventional and advanced effluent treatment, in environmental processes and finally during drinking water treatment. The majority of pharmaceuticals taken by a patient will be at least partially excreted in the form of metabolites or conjugates. Often degradation in sewage and water treatment and the environment is incomplete resulting in stable transformation products. The formation and presence of such stable transformation products in the effluent of sewage works, surface water, and drinking water treatment is reported in the scientific literature with increasing frequency (see for example Kosjek and Heath 2008 , Perez and Barcelo 2007 , Gröning et al. 2007 , Vasconcelos et al. 2009 , Längin et al. 2009 ).
This is even more of importance as advanced oxidation techniques employing e.g. ozone, hydrogen peroxide, light or electro-coagulation are increasingly under discussion for the removal of pharmaceuticals and other micro pollutants in effluent treatment and drinking water treatment. In general the literature dealing with TPs is “exploding”. Because of non-standardized approaches, the number of possible treatment technologies and individual test conditions it is difficult to keep an overview on all work published for pharmaceuticals in general.
Toxic compounds may be generated within treatment formed. In some cases the transformation products can be even more toxic, e.g. genotoxic, than the parent compound (e.g. Isidori et al. 2005 , 2007; Lee et al. 2007 ; Wei-Hsiang and Young, 2008 ). Through the application of analytical techniques for structure elucidation, the chemical structure of such compounds can be established. Unfortunately, for most transformation products only little information on their properties is available for two reasons: 1) They are not available in amounts necessary for experimental testing as isolation and/or targeted synthesis is too expensive and 2) the number of compounds described in literature is too high to be tested for reasons of time and cost. Therefore, data for a full and reliable risk assessment of parent compounds and their resulting transformation products is missing.
A solution for this undesired situation is offered by the employment of computer based methods (“in-silico screening and assessment”, for an overview see Ekins 2007 ). Using in silico screening tools only the structural formula is necessary to identify the compounds of highest interest with regard to their physico-chemical and toxicological properties.
In this context, the project aimed
• to collect knowledge on stable transformation products of antibiotic and antineoplastic compounds (parent compounds selected within in the PHARMAS project) and assess experimental approaches by an extensive literature review,
• to study experimentally the fate of parent compounds selected within in the PHARMAS project and TPs generated different processes proposed for water and waste water treatment and environmental processes (effectiveness and treatment times, identification of TPs and assessment of the toxicity of TPs using experimental and in-silico methods),
• and to develop a first basis for the targeted design of pharmaceuticals with enhanced degradability and elimination behaviour (“benign by design”).
An extensive literature search was conducted in the scientific databases SciFinder (database of CAS), Web of Knowledge, and PubMed. The aim of this part of work was to investigate whether TPs of the target compounds selected within the PHARMAS project are formed during the different types of treatment processes for effluents and water and to collect information on their chemical structure, respectively. The treatment processes, which might lead to transformation products, were: photolysis, biodegradation, chlorination, ozonation and other advanced oxidation processes. Results of additional studies, in which more complex treatment processes were investigated, were also taken into account, e.g. formation of TPs during combinations of the above mentioned ones (e.g. chlorination and ozonation) ozonation combined with biological reactors or degradation studies in outdoor microcosms which simulated complex natural water environment.
The literature study shows that the knowledge on transformation products formed is scarce despite a steep increase of publications dealing with this issue in recent years. However they focus most often on a narrow spectrum of compounds using different treatment conditions (details often not sufficiently detailed provided). Out of all selected pharmaceuticals of PHARMAS for four, namely imatinib, capecitabine, hydroxycarbamide and ifosfamide, there was completely no information available. Interestingly, this is a mixture of old and new drugs. For others, such as ciprofloxacin much more than fifty different TPs were described in literature relating in some cases to different conditions of there formation. Studies focusing on the transformation products of pharmaceuticals formed by chlorination are currently only few for all compounds. No reports were found which have dealt with this topic for selected cytostatic drugs. In general there is more information available on TPs formed from antibiotics than from cytostatic drugs. The largest part of older studies focused on the investigation of the elimination kinetics of parent compounds without presenting any information on the chemical structure of individual TPs formed and their amount formed. In more recent publications there is an increasing trend to deliver such information.
In a minority of the publications dealing with the compounds that were of interest within the Pharmas project there was information presented which was related to the identification of TPs formed and/or on the assessment of the biological potency of them when present in mixtures with parent compound. In most of the cases TPs were “identified” by using most often mass-spectrometry analysers such as LC-MS/MS, MSn, a few using more advanced high resolution mass spectrometry. Based on the MS results chemical structures were suggested. The reliability of such proposed structures of TPs could not be assessed in all cases on the basis of the information presented. On the other hand there were also studies that focused on appearance of TPs based on comparison with analytical reference standards if available at all (e.g. human metabolites). So far there were only few lab based studies on the further fate or effects of single TPs. In most of the cases, studies focussed on structure elucidation and pathways of TPs formation. Some of the studies assessed additionally the toxicity of mixtures of TPs formed after a certain time of treatment or biodegradability of this kind of mixtures. Due to cytotoxic and genotoxic behaviour of cytostatic drugs an impact on the environment could be expected but was not assessed. There is usually a decrease of antibiotic activity among TPs formed from antibiotics. However there are examples of increased toxicity of those TPs or no significant difference in toxicity between TPs and parent compounds. The details of the review are collected in the deliverable 5.1.
In order to enhance quality of drinking water and to eliminate pollutants in sewage water, several techniques have been employed for many years. With emerging pollutants, new techniques have come into focus. Treatment steps for both kind of water include generally (depending of the size of the plants) retention (filtration, flocculation, sedimentation) coupled to biodegradation and advanced processes (oxidation by different chemicals, disinfection, photolysis. Retention processes consist of the transfer of the substances of interest from the liquid to a solid phase and the elimination of this last one. Contrarily, degradation processes conducted either result in the total elimination of the substances (mineralization if no sorption) or to the incomplete transformation resulting in products that need to be characterized (in particular in terms of toxicity) or are totally inefficient. The outcome of such treatments depends heavily on the compound and the type of the applied treatment. A number of active ingredients of pharmaceuticals have been found to be persistent in regard of wastewater treatment as well as drinking water treatment. Thus, although treatment of drinking water and sewage water is well established, there are still many treatment-resistant substances present. Thus the improvement of the efficacy of the treatment processes leads to the formation of transformation products that can pose a risk for the environment and the public health.
Based on the above mentioned results, the chemical behaviours of a selection of antibiotics and cytostatic drugs in selected sewage and drinking water treatment processes were investigated obeying the following scheme (Figure 9).
Figure 9: General approach for the investigation of the fate and biological effects of TPs the selected pharmaceuticals
For the further investigation, a selection of pharmaceuticals of both classes was done based on literature review. According to a recent review, 5-fluorouracil, cyclophosphamide and methotrexate were the most widely consumed anticancer drugs worldwide. Imatinib belongs in contrast to the above mentioned group of newer anticancer drugs on which nearly no information was available. Ciprofloxacin is a heavily used antibiotic that was shown not to be biodegradable in sewage treatment in several studies. The cytotoxics methotrexate, cyclophosphamide, imatinib and 5-fluorouracile, and the antibiotic ciprofloxacin were therefore investigated in laboratory and pilot scale studies. Slightly different approaches were developed for treatment of drinking water and sewage/sewage plant effluent (see Error! Reference source not found.9) to address the differences related to drinking water on the one hand and waste water on the other one. In total, for the five pharmaceuticals investigated in this study, eight different treatments from drinking water treatment and sewage water treatment were tested. All pharmaceuticals gave transformation products in at least in one treatment; for most of these, no finally reliable structure could be proposed. The following treatments were investigated: LI (low intensity) and HI (high intensity) UV-photolysis, Xenon photolysis, chlorination, ozonation, and a combination of the latter two.
5-FU, imatinib and related cytotoxcis were determined in the water samples. As we could only find TPs for the treated imatinib and CPA the samples were only run with those settings (positive ionization and PFP and HILIC LC columns respectively). Considering the low levels of IB found in the water samples (<5 ng/L) it is not likely to detect the TP estimated to constitute ca 5% of the formed degradation products in the water samples, assuming a similar LOD (5ng/mL). Also, neither imatinib related TPs nor CPA TPs could be detected. A clean up method targeting the TPs specifically would be necessary to be able to determine their presence and concentration.
5-FU was fully degraded by chlorination and ozonation. After both treatments the formation of TPs has occurred (as evidenced by DOC monitoring), whereas chlorination after ozonation degrades the formed TPs. 5-FU was not affected by xenon lamp or biodegraded, but UV fully degraded 5-FU. TPs were also expected to be seen after UV lamp treatment. Probably due to the small molecular size of this anti-cancer drug, and the polarity of the fractions, no masses could be detected and isolated in with the analytical detect, neither in positive nor in negative ionization mode. This example demonstrates that in some cases for TP identification other strategies need to be considered.
CPA was not affected by chlorination or Xenon lamp treatment, which indicates that there is no full mineralization of the compound. Hence, TPs could be expected (confirmed also by DOC measurements). It was degraded (ca. 75%) by ozonation, UV lamp and biodegradation. In other words, none of the treatments fully removed CPA. Two CPA TPs were found, one after ozonation and one after biodegradation.
Imatinib was fully degraded with chlorine and ozonation, partially by UV lamp treatment but not by biodegradation. The DOC measured after chlorination, ozonation and UV lamp treatment indicated TP formations.
Regarding ciprofloxacin and methotrexate, structures for transformation products resulting from chlorination and UV-photolysis could be proposed.
Parent compounds and the mixtures resulting from LI (low intensity) and HI (high intensity) UV-photolysis, Xenon photolysis, chlorination, ozonation, and combination of the latter two were tested for their toxicity with the AMES Aqua test. Toxic effects were not observed in any of these scenarios. Even the parent compounds (with the exception of ciprofloxacin) were not toxic at the concentrations applied. The tested concentrations did not pose an acute toxic potential within the tested parameters of the AMES aqua test.
TPs were additionally tested with quantitative structure activity relationships (QSAR) regarding various additional toxicological endpoints if an unambiguous chemical structure was available. Regarding methotrexate, chlorination did not result in less toxic transformation product. One transformation product formed during LIUV-photolysis had even more alerts as the parent compound thereby giving hints of increased toxicity through treatment processes. Proposed transformation products for ciprofloxacin possess mostly the same alerts as the parent compound. Here, no change in the pattern of the toxicity could be found. One transformation product (chlorination M 2_2) gave much less alerts as the other.
Summarizing treatment processes, all investigated pharmaceuticals formed at least during one treatment transformation products. Most of them are not known. Further work is needed to confirm the identity of the TPs. Drinking water treatment processes including chlorination lead preferably to transformation products that contained an atom of chlorine which is not first hand to be expected within photolysis treatment (corresponding to sewage water treatment processes). A reason could be that by photolysis active chlorine containing species are generated from inorganic chloride. As most of the transformation products could neither be identified nor fully assessed regarding toxicity, further research is needed. In some cases structures could be put forward but distinction of different isomers was not possible and thereby preventing QSAR analysis.
The results have been collected in deliverable 5.2.
By applying computer-based methods such as QSAR it could be demonstrated that relevant alerts for non-degradability could be identified and neutralized in order to construct a sustainable version of the parent compound. This strategy was applied to ciprofloxacin and its transformation products, but evidence was found that it is most likely also applicable to other substances. Furthermore, molecular docking software could help to provide a first indication that the altered compound might still be pharmacologically active without the need to synthesize the molecule beforehand and subsequently to test it in the wet lab (deliverable 5.3). Candidates which pass this in-silico validation procedure can be selected for further investigations and might possess opportunities to generate alternative or improved pharmaceuticals. In this respect the applicability and advantage of in-silico tools as a step in the direction to benign by design pharmaceuticals of the future was shown. Some first guiding principles were identified: With ciprofloxacin as an example for constructing environmentally improved pharmaceuticals by computer methods (so-called “benign by design” process) was specifically demonstrated. This approach will pave the way to developing practicable ways to design degradable pharmaceuticals (see deliverable 5.4).
C.5. Risk assessment
A specific task was dedicated to risk assessment with the aim to integrate the information concerning exposure, effect of single and mixture products in order to estimate the environmental risk posed by the use of the selected pharmaceuticals for a selected number of endpoints. More specifically, the main objectives were:
1. Set up realistic worst case exposure scenarios for exposure of sensitive subgroups of the human population to a selected number of pharmaceuticals;
2. Characterize the cumulative risk due to exposure of humans and non human endpoints (aquatic ecosystems, and wildlife) to a selected number of pharmaceuticals;
3. Assess the uncertainty and variability involved in the risk characterization by means of probabilistic techniques and expert assessment of uncertainty;
4. Explore alternative risk indicators which are more informative than the conventional PEC/PNEC ratio.
To reach these objectives, a screening tool was developed to identify combinations of human pharmaceuticals, locations and exposure groups in Europe for prioritization (deliverable 4.1). The screening tool was adapted into a location-specific decision support tool to be used by physicians for the prescription of drugs. Furthermore, an extensive uncertainty analysis was performed. At first, this analysis focused on quantifying uncertainty in the acceptable daily intakes of two pharmaceuticals, i.e. ciprofloxacin and methotrexate. Subsequently, uncertainty in human and ecological exposure estimates was quantified by means of Monte Carlo simulation. An expert workshop was organized to assess the sources of uncertainty that cannot be quantified in a Monte Carlo analysis. Finally, a study was performed to optimize the use of information in ecological risk assessment of pharmaceuticals using the multi substance Potentially Affected Fraction (msPAF) as an endpoint.
The main outputs are resumed below.
A European wide, spatially explicit screening tool was developed that was used to identify locations in Europe showing the highest human and ecological risks due to the use and presence of 11 selected human antibiotics and 7 antineoplastics.
For human risk assessment, a range of exposure scenarios was defined, i.e. based on age (0-1 years, 1-10 years, 18-65 years and >65 years), drinking water purification technique (none, conventional and advanced), diving behaviour (none, recreational and professional; 18-65 years only), and food origin (local and Europe-wide). Human contact media that were taken into account were drinking water, fruits and vegetables, meat products, milk products, fish, surface water and soil. HD50 and HC50 values were used to characterize the human and aquatic toxicities of the pharmaceuticals, respectively. These toxicity values represent the hazardous dose (HD50) or hazardous concentration (HC50) of a chemical at which at least half of the individuals in half of the species is affected. HD50 and HC50 values were based on acute toxicity data, due to data availability and under the assumption that relative acute and chronic toxicities of the substances are similar. The risk quotients for all combinations of substance, location and exposure group were calculated by division of concentrations (for the aquatic environment) or daily intakes (for human health) by the toxicity values. Finally, summed risk quotients for all antibiotics and for all antineoplastics were calculated per grid and exposure group (for human health) on the basis of the concentration addition principle. Figure 10 shows that the summed aquatic risk quotients due to exposure to antibiotics are substantially larger than those due to exposure to antineoplastics, a pattern that is also visible at the level of the individual compounds.
Figure 10: Summed aquatic risk quotients (predicted surface water concentrations / HC50) due to exposure to (A) antibiotics and (B) antineoplastics. The grids with the highest summed risk quotients are outlined.
The highest aquatic risk quotients are calculated for three antibiotics: levofloxacin, doxycycline and ciprofloxacin. This can be explained from their aquatic toxicities (i.e. second, third and fourth most toxic of all compounds assessed), as well as their very low removal in both STP and the environment. Risk quotients were highest in the Milan region in northern Italy. This can be explained from the high levofloxacin consumption in Italy, compared with the other EU Member States, and the high population density of the region.
Highest human health risk quotients are estimated for infants that consume conventionally treated drinking water. Figure 11 shows that the difference between the summed health risk quotients due to exposure to antibiotics and due to exposure to antineoplastics is not as clear as for the aquatic environment (Figure 10).
Figure 11: Summed health risk quotients (predicted exposure / HD50) for infants due to exposure to antibiotics and antineoplastics, after consumption of conventionally treated drinking water and either locally produced or Europe-wide food. A = antibiotics and locally produced food; B = antineoplastics and locally produced food. The grids with the highest summed risk quotients are outlined.
There is a large influence of food origin on the health risk quotients. 5-Fluorouracil and to a lesser extent ciprofloxacin have the highest health risk quotients when a local food origin is considered. In the two grids located in eastern Spain (Figure 11), sludge application on agricultural soils is very high and run-off from agricultural soils is expected to be rather low.
The tool is to our knowledge the first that can be used to prioritize combinations of human pharmaceuticals, locations and exposure groups, in such a way that potential hot-spots and substances can be identified for future monitoring activities and future research in the framework of aquatic and human health risk assessment. All the details have been reported in deliverable 4.1 and have been published in an international scientific journal
In a second task, we explored alternative endpoints for human and ecological risk assessment. In order to realize this aim, the screening tool (described above) was adapted to make it suitable for the assessment and comparison of the environmental impact of two alternative pharmaceutical prescriptions. This methodology provides physicians with the opportunity to include environmental considerations in their choice of prescription. A case study with the two antibiotics ciprofloxacin and levofloxacin at three locations throughout Europe showed that the preference for a pharmaceutical might show spatial variation, i.e. comparison of two pharmaceuticals might yield different results when prescribed at different locations (Figure 12). This holds when the comparison is based on both the impact on the aquatic environment and the impact on human health. The relative impacts of ciprofloxacin and levofloxacin on human health were largely determined by the local handling of secondary sludge, agricultural disposal practices, the extent of secondary sewage treatment, and local food consumption patterns. The relative impacts of ciprofloxacin and levofloxacin on the aquatic environment were mostly explained by the presence of specific sewage treatment techniques, as effluents from sewage treatment plants (STPs) are the most relevant emission pathway for the aquatic environment.
Figure 12: Relative impact of ciprofloxacin and levofloxacin on the aquatic environment (Aq.) and human health (Hum.), in northern Italy (left), south-eastern Spain (middle) and eastern Sweden (right). Blue bars: relative impact of the prescription of one daily defined dose (DDD) of ciprofloxacin; Yellow bars: relative impact of the prescription of one DDD of levofloxacin.
The use of alternative risk assessment endpoints was furthermore addressed in a study that focused on ecological risk assessment of pharmaceuticals. The main aim of this study was to optimize the use of available ecotoxicological data to assess the risk of multiple pharmaceuticals to aquatic ecosystems. In this study, the use of the multisubstance Potentially Affected Fraction (msPAF) was used as an alternative endpoint to quantify the impact of multiple substance on aquatic ecosystems. A combination of concentration addition, response addition and toxic pressure addition was used to predict the combined impact of multiple substances. Bayesian techniques were used to combine substance-specific ecotoxicological data with other sources of information contained in ecotoxicological databases, thereby reducing the uncertainty in the msPAF. The analysis was performed for a realistic mixture of pharmaceuticals in the German Ruhr area, as predicted by the prioritisation tool outlined in D4.1. The study is described in detail in Deliverable 4.4.
Finally, a detailed and integrated probabilistic risk assessment for a selected number of high risk exposure scenarios proposed. Several studies and activities were performed to realize this aim. First, the screening tool for the location-specific prioritization of human pharmaceuticals in Europe was subjected to a probabilistic assessment with Monte Carlo techniques. This way, the uncertainty underlying the calculations was quantified. Uncertainties were analysed in an integrated way; combining ecological and human risk assessment where possible. Next to a complete assessment of the whole chain of calculations from consumption data to risk quotients, the tool was divided into six sub models, which were then separately assessed. These six sub models were 1) Emissions to wastewater, 2) STP removal, 3) Environmental fate, 4) Drinking water purification, 5) Human exposure, and 6) Toxicity and effect. Three substances were selected for a case study, being ciprofloxacin, erythromycin and 5-fluorouracil. They were chosen because they represent both antibiotic (ciprofloxacin and erythromycin) and antineoplastic substances (5-fluorouracil), and because they differ in their ionizing properties, i.e. ciprofloxacin is a zwitterionic compound, erythromycin is a base, and 5-fluorouracil is an acid. Furthermore, three STPs in three different grids were selected for the case study. This selection was based on the range in treatment techniques applied between the three STPs, and the variation in the extent to which secondary sludge is disposed of between the three Member States of concern. Parameters were assigned a (log-)normal distribution, preferably based on literature data or, if data were insufficient, based on related data. An exception was made for parameters that represent fractions, which were characterized with a beta-distribution, based on literature data. Parameters for which a quantification of the uncertainty was not possible on the basis of literature data were assigned an uncertainty distribution based on expert judgement. The results of this study are described in detail in Deliverable 4.2.
Second, a study was performed to separately quantify different sources of uncertainty and variability in the derivation of the Acceptable Daily Intake (ADI) of two pharmaceuticals, i.e. ciprofloxacin and methotrexate. A separate propagation of all sources of true uncertainty and interindividual variability throughout the derivation process enables policy makers to set separate standards for the fraction of the population to be protected and the confidence level of the exposure limit protecting that fraction. Additionally, the methodology enables the inclusion of potentially sensitive subpopulations, and it identifies the main sources of uncertainty and variability influencing the exposure limit, therefore providing a focus for future research. For both substances studied, the ADI was mainly influenced by uncertainty in the derivation of a point of departure, the extrapolation from sub-chronic to chronic toxicity, and by uncertainty in the extrapolation from mammalian to human toxicity. It was shown that traditionally derived ADIs protect 95%-99% of the population with 95%-99% confidence. The framework for deriving probabilistic ADIs is illustrated in Figure 13. The results of this study are described in detail in Deliverable 4.3.
Figure 13: The framework for the probabilistic derivation of ADIs (U = uncertainty; IV = interindividual variability; S = systematic difference).
Third, an expert workshop entitled Uncertainties in Human and Ecological Risk Assessment of Human Pharmaceuticals was organized on 24-25 October 2013 in Ravenstein, The Netherlands. The main aim of the workshop was to identify, describe and assess the uncertainties currently involved in risk assessment of human pharmaceuticals. This information is particularly important for planning future research to further optimize the risk assessment process. The 26 experts that participated in the workshop were divided over 6 working groups, i.e. (1) emissions to wastewater, (2) STP removal, (3) environmental fate, (4) drinking water purification, (5) human exposure, and (6) toxicity and effects. Each working group identified, described and assessed the uncertainties within their particular theme, using the screening tool as a starting point for the discussions. The assessment and prioritization of the uncertainties proceeded according to an established methodology which was supervised by an external expert, i.e. dr. Anne Knol. The workshop provided detailed insight in the main sources of uncertainty involved in assessing the human and ecological risks of human pharmaceuticals. A detailed analysis of the results is ongoing and will be processed into a joint scientific publication of the workshop participants that is expected by the end of 2014.
Potential Impact:
A European risk and hazard classification system for medicinal products would be a strong core element of precautionary EU policies for minimising possible risks for human health and ecosystems due to environmental exposure to pharmaceuticals and their derivatives. Such a system would primarily be designed as a (web-based) decision support tool for practitioners (e.g. physicians and pharmacists) providing straightforward action alternatives. There is initial evidence from a five year experience with such a system in Sweden that it can influence prescription, sales and consumption patterns of pharmaceuticals in terms of promoting less problematic substances where therapeutically equivalent alternatives are available (LIF, 2007). Based on the Swedish experience, there have been on-going discussions on the possibility of extending t his scheme to the rest of the EU. At the same time, concerns raised in the completed EU-funded project KNAPPE indicated that the lack of available data on the effectiveness of the scheme could be a barrier to its implementation in other countries. Some stakeholders see a thorough evaluation of the experiences with the Swedish “pilot” system as a prerequisite for its extension. The stakeholder community for such a system is most certainly a wider group than just practitioners and pharmacists; data on pharmaceuticals may also serve as aid in decision-making for drinking water plant managers, river basin managers, managers of waste water treatment plants etc.
In this context, and in addition to the scientific and technical works described above, another task more focussed on stakeholders and society was performed. The main aims were
- to explore stakeholder requirements on an EU wide system, to assess its influence on prescription, sales and consumption patterns of selected pharmaceuticals, to estimate its socio-economic impact, and, based on these findings, to develop a prototype tool that in particular accounts for the different markets, health care systems and cultural background of EU member states.
- to identify stakeholders' level of information and understanding of the issue, as well as associated risk conceptions; this will help define the characteristics required of possible risk communication alternatives, as well as influence project dissemination.
To answer these objectives, three main tasks have been implemented: (i) a qualitative socio-empirical study on the specific requirements of the European stakeholder community regarding such a classification system; (ii) a Delphi-based socio-economic impact assessment of an European classification system using the example of three countries and comparison to the Swedish system; (iii) a prototype database for risk and hazard classification of pharmaceuticals.
For the qualitative socio-empirical study on the requirements of different stakeholders on a risk and hazard classification system for pharmaceuticals in the environment, a short inventory of the specific European stakeholder community was set up as a first step and the perspective of stakeholders on the issue was explored via in-depth phone interviews with a group of key stakeholders. The questions focused on:
• the stakeholders’ attitude towards such a possible system,
• their evaluation of its use and of possible impacts,
• their information requirements from such a system,
• their own use of such a system (e.g. in their decision-making processes),
• their opinions on the characteristics and design it should have, and
• their risk perceptions on the issue of pharmaceuticals in the environment
The review of stakeholders’ requirements showed two main concepts of what such an information system should contain, and what it should deliver. These have been termed “knowledge base” and “decision support system for doctors / pharmacists / patients”. These concepts do not exclude each other, nor do they have to be independent of each other. Several stakeholders suggested the need for a system that covers both approaches. Key recommendations included:
• An information system on PIE should not be exclusively a decision support system for doctors/pharmacists/practitioners, but also have elements of a knowledge base approach. The stronger the knowledge base component of such an information system, the higher the chances for interesting present and future uses for the information and for positive impacts on the environment.
• Data should not only be complex and aggregated. Simple, “raw” data should also be made available, so as to fulfill the requirements of stakeholder groups beyond the medical system.
• In view of the data and knowledge gaps on the topic and of the importance of precaution in the water cycle, a more ambitious approach to an information system on PIE seems preferable. An ambitious information system on PIE would however require a certain level of resources.
• An independent body in charge of an information system on PIE has strong support among stakeholders, and is seen as key to generate trust in the system. Due to the high complexity of this information and the possibility of errors affecting this trust, data should not just be reviewed for internal consistency, but should be extensively proofed and compared with other available data.
The overall details of the study have been reported in the deliverable D6.1.
A second task was focused on the possible future impact of an European Classification System and the comparison with the Swedish system. First, five qualitative interviews with experts from the pharmaceutical industry, authorities (Stockholm County Council and Swedish Association of Local Authorities and Regions) and science have been conducted in summer 2011. The result was: there is not sufficient quantitative data for an impact assessment of the Swedish classification system.
Other results were that the classification does not seem to be used extensively by individual doctors, mainly because it seems to be not very user friendly. On the other hand, it was thought that the fass.se system is a first and important step to collecting information on the topic and it helps to raise awareness around the issue.
The real impacts are rather expected to occur through the “wise lists”, as county councils increasingly use the information from the classification system to compile those lists, which doctors often consider in their prescription decisions.
A problem seems to be the lack of adequate data. At the time of the interviews, 50% of all substances were not classified. So it is not possible to take Sweden as a quantitative benchmark and to compare it with the results of the following Delphi-Study.
The Delphi method is a structured survey and communication technique, developed as a systematic, interactive forecasting method which relies on a panel of experts. In the Pharmas-Project it was conducted in Germany, the United Kingdom and Hungary. In total, responses were obtained from 186 experts, from various different stakeholder groups which are relevant for the issue of pharmaceuticals in the environment. Besides practitioners, pharmacists, patients and the pharmaceuticals industry, they represent water utilities, medical, pharmaceutical and environmental authorities, NGOs and scientists.
Analysis of the survey responses showed that there is a very high approval of the introduction of an EU-wide environmental classification system for pharmaceuticals. However compared to other possible measures, such as intensified water treatment or proper disposal of unused pharmaceuticals, it is not a top priority. Furthermore, experts assume that the classification system will be used by a significant share of relevant stakeholders (for example between 17 and 59% of doctors depending on the country and framework conditions) and that significant reductions of pharmaceutical residues in the water are possible (between 23 and 50% depending on the country and the group of pharmaceuticals). The results of the first round of the questionnaire have been presented in deliverable 6.2.
Finally, applying the results from the studies described above, and by exploiting the experiences with the existing Swedish system, a prototype for an EU-wide risk and hazard classification system for human pharmaceuticals has been developed. The core of the system is a database that comprises and discerns relevant data on cytostatics and antibiotics. The overall result from Tasks 6.1 and 6.2 on the prerequisite of an EU-wide risk and hazard classification system was the requirement of a knowledge based decision support system (DSS) with varying levels of details regarding the environmental risk of the active pharmaceutical ingredient (API). From a simple hazardous phrase to extensive detailed information, such as sales data, environmental distribution, toxicity, biodegradation, bioaccumulation, removal in sewage treatment plants (STPs), chemical and physical properties of APIs. A common statement from stakeholders was that the more solid and strong the foundation of the knowledge base component becomes - the higher the chances for interesting present and future uses for the information and for positive impacts on the environment. An additional requirement from several stakeholder groups was that such in-depth information should be made available to view and to extract in a simple way. The prototype is described in Deliverable 6.3 and is available at http://eupharmas.ivl.se/
Communication
Project communication and result dissemination were addressed in an individual work package (WP7). This section summarises the actions and results of WP7 over the project’s 39 months.
“Pharmaceuticals in the environment” is an issue receiving growing attention. In particular, information regarding the effect on organisms is more and more required by the actors with influence on the life cycle of PPs, which includes regulators and the general public.
The objective of this work package was to expand and disseminate, as widely as possible, the main information on the many dimensions of this topic. In other words, the aims were to:
• increase the visibility of the topic among the scientific and the policy-making communities by fostering communication and collaboration between scientists and regulators, so as to catalyse and promote further exploration, discussion and collaboration on PPs by all stakeholders and beneficiaries of the work.
• To reach out to the general public (as a group with strong possibilities to influence the presence of pharmaceuticals in the environment) by attractive, simple and easy-to-read language.
• To target media by preparing specific supports corresponding to their requirement and to facilitate their distribution
Communication efforts were addressed to a wide range of users (scientists, the general public and media, and policy makers) and promoted in different ways: website and multimedia support (CDRom, interactive web animations), newsletters and leaflet, a policy brief, and events (corresponding to the 5 sub-activities of the WP).
Task 7.1 concerned the development and maintenance of a dedicated website. The structure was discussed among the partners; the actual website implementation was subcontracted by the project’s scientific coordinator. Ecologic Institute developed texts to address the general public and policy-makers. The web site is accessible at the following address: www.pharmas-eu.net. Figure 13 illustrates the website home page. The main structure of the site has been presented and described in Deliverable 7.5.
Figure 13: Pharmas website home page
Task 7.2 was dedicated to the production of printed material: project newsletters (targeting the scientific community and policy makers) and leaflets (targeting the broad public).
A total of 6 project newsletters were developed over the project duration. Armines held the lead responsibility for this tasked, and was assisted by EHESP and Ecologic Institute. The newsletters were 4 pages in A5 size, and were made available for download on the project website. Starting with Newsletter 3, newsletters were also distributed to an email list of over 700 scientists and policy stakeholders involved in the topic of pharmaceuticals in the environment. All NLs were also made available on the project website for download.
NL1 introduced the project, its aims, and the project consortium. The following NLs presented simplified summaries of the scientific results of different work packages/tasks. The release dates of the different NLs was adjusted from the original plan (planned every 6 months), according to the availability of project deliverables and relevant themes. For instance, NL5 was timed so as to summarise the results of the project-organised International Conference on Pharmaceuticals in the Environment, held in early June 2013 in Nimes, France.
A leaflet targeting the broad public was developed in the final project year. At the 3rd Project Meeting (February 2013, Copenhagen) the decision was taken to develop a leaflet targeting the public that is not focused on the project or on project results, but rather on providing simple but accurate information on the topic of pharmaceuticals in the environment. Ecologic Institute drafted a first version of this leaflet, which was distributed to project partners, who provided valuable comments and suggestions to improve the message and its delivery. Finally, Armines and Ecologic Institute translated the English original into French and German. The leaflet will be permanently available over the CD-Rom, has been made available on the project website, and circulated to the over 700 contacts in the project distribution list.
Project newsletters and leaflets are collected in Deliverable 7.1.
Task 7.3 had as aim informing policy-makers and policy stakeholders of project results, by presenting them at meetings, organising events, and using policy briefs and policy-relevant communication channels.
Under WP7 of Pharmas, Ecologic Institute organised a Science-Policy Event in November 2013, which took place in Brussels. The event presented the results of 5 projects on pharmaceuticals in the environment: 2 FP7 projects (PHARMAS and CYTOTHREAT), an Interreg project (noPills), a German Environmental Protection Agency project (Global Relevance of Pharmaceuticals in the Environment), and a UN project (WHO project on greening procurement of pharmaceuticals). DG Research provided the facilities for the event, which was attended by approximately 30 policy stakeholders, active at EU or at Member State level. The documentation of this event (slides, presentation summaries, stakeholder position statements) is available on the project website.
In March 2014, the main results of the PHARMAS project were presented in a meeting of the Working Group E (Chemicals) of the Water Framework Directive’s Common Implementation Strategy. Representatives of the European Commission and of the 27 Member States were given a summary of the final project results.
In addition to the presentation of project results at purely scientific events (which members of WPs 1 to 5 carried out), presentations on PHARMAS were also held at conferences and events which partly attract policy stakeholders, such as the Berlin Water Week 2013, and the 3rd International Conference on Sustainable Pharmacy 2012. Thomas Backhaus (WP3) presented the project at the 5th EUFEPS World Conference, on Drug Absorption, Transport and Delivery, a conference attended mainly by people from pharmaceutical industry. Ecologic Institute presented and disseminated information on the Pharmas project to visitors (policy-makers, scientists and stakeholders) of the 12th edition of Green Week, the biggest annual conference on European environmental policy, which took place on 22 - 25 May 2012 in Brussels (the focus for 2012 water, and the Green Week was titled "Every Drop Counts").
Partners of other WPs were also encouraged to use policy-relevant channels of communication to disseminate their scientific results. The article of Rik Oldenkamp et al. (2013) (WP4) was turned in and featured in the Science for Environmental Policy Newsletter (Issue 332), which enjoys wide readership in policy circles.
The results of the different PHARMAS work packages have been summarised and simplified, so that they are accessible for policy stakeholders. This policy-relevant summary of project results have been made available in the project’s Policy Brief. Dissemination of the Policy Brief is through project website and distribution over the project’s electronic distribution list.
The Policy Brief is presented in Deliverable 7.2.
Task 7.4 concerned the organisation and execution of an International Conference on Pharmaceuticals in the Environment. This event was initially planned for December 2012. After discussions with the Cytothreat coordinators, it was decided to hold a common conference with results from both projects in June 2013 in Nîmes. The Conference was titled “Pharmaceutical Products in the Environment: Is there a Problem?”, and gathered 110 participants coming from USA, Canada and Europe. Eastern European countries were strongly represented thanks to CYTOTHREAT partners and their networks.
A keynote lecture, two plenary lectures, 24 oral presentations and 35 posters were presented during the two days conference. It was split into 3 sessions:
1. Actions at the European level
2. Environmental and Human Risk Assessment of Pharmaceuticals
3. Perspectives on Problem Assessment
Backdrop to the discussions was the substantial amount of work on the topic carried out during the last 15 years. Over this period, the scientific, regulatory and industry communities have performed research on occurrence, fate, effects and risks of pharmaceuticals in the environment. Regulations have also been developed on the risk assessment of environmental exposure to these compounds. The conference focused on the following questions: What are the lessons learned from all this new knowledge? Can they guide the development of future research and policy initiatives? And is there a real problem with pharmaceuticals in the environment?
Task 7.5 covered the use of multimedia to disseminate the project topic and project results.
Under this task, Ecologic Institute developed a web animation on the topic of pharmaceuticals in the environment. This was the main means to reach out to the general public on this topic. The animation was produced between September 2012 and May 2013, and publicly launched at the above-mentioned International Conference on Pharmaceuticals in the Environment, held in Nimes on June 2-3, 2013.
The idea to use a web animation to communicate the science on the topic to the broad public was due to the possibility of using graphic illustrations to visualise issues that are otherwise hard to visualise, the possibility of attractive visual and sound effects, as well as the possibility of communicating the topic in a didactic and entertaining way. The animation’s storyline has the following structure:
1. Take up some common myths about the impact of pharmaceuticals on human health, poke some fun at them.
2. Dismiss these as not scientifically proven, current scientific consensus is that this subject poses no risk to human health.
3. However, there are instances of significant impacts of drugs on the environment: mention examples.
4. Explain how pharmaceuticals make their way into the environment.
5. Highlight actions being taken by science and the policy world.
6. Highlight actions that consumers can take to reduce their own input of pharmaceuticals into the environment.
The animation’s full title is “The Drugs We Wash Away: Pharmaceuticals, Drinking Water, and the Environment”. It can be viewed on YouTube under the following link: http://www.youtube.com/watch?v=OYbRlJLBzn4 . The INTERREG project “noPills” and its partner project “Den Spurenstoffen auf der Spur” (DSADS), working on the same topic in Germany, produced a German version of the animation. This version is available under: http://www.youtube.com/watch?v=6blafjHQGvw .
At the moment of writing, the English version has 9,506 views, whereas the German one has 1,011 views, thus summarising over 10,000 clicks, a very good result for this type of animation.
Furthermore, noPills and DSADS have developed a follow-up animation, using the same animation studio, visual and sound styles, and characters of “The Drugs We Wash Away”. This animation, also available in English and German, was launched at the end of March 2014, and is available under: http://www.youtube.com/watch?v=Y_-DBrrCDA0 and http://www.youtube.com/watch?v=zi_qfXxyEyI .
A detailed report on the animation and its production process is given in Deliverable 7.3.
A CD-Rom was produced to help disseminate the scientific results and ensure the legacy of the project. The CD-Rom follows the format of a webpage (its Start Page opens in a web browser), and provides a complete overview of project, project objectives, project structure, results of the events organised within the project, project consortium, as well as including all project deliverables. The navigation and structure are intuitive and easy to follow.
The CD-Rom is Deliverable 7.4.
List of Websites:
www.pharmas-eu.net